Verbalizable Representations Form a Global Workspace in Language Models

Adam Pearce; Anna Soligo; Ben Thompson; David Abrahams; Emmanuel Ameisen; Euan Ong; Isaac Kauvar; Jack Lindsey; Joshua Batson; Mateusz Piotrowski

REVIEW 4 major objections 5 minor 8 cited by

Verbalizable Representations Form a Global Workspace in Language Models

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Language models maintain a small, privileged set of internal representations—called the J-space—that act as a global workspace: reportable, deliberately controllable, and used for silent reasoning while automatic processing proceeds without

desk verdict The Jacobian lens is a real new technique with strong causal controls, but the workspace claim leans on a circular verbal-report pillar and several post-hoc choices; deserves serious refereeing, not desk rejection. read the letter →

arxiv 2607.15495 v1 pith:4SYZYYTN submitted 2026-07-16 cs.CL cs.AIcs.LG

Wes Gurnee , Nicholas Sofroniew , Adam Pearce , Mateusz Piotrowski , Isaac Kauvar , Runjin Chen , Anna Soligo , Paul Bogdan

show 8 more authors

Euan Ong Rowan Wang Ben Thompson David Abrahams Subhash Kantamneni Emmanuel Ameisen Joshua Batson Jack Lindsey

This is my paper · ORCID

classification cs.CLcs.AIcs.LG

keywords JacobianlensJ-spaceglobalworkspaceverbalreportinterpretabilityalignmentauditingcounterfactualreflectiontraininglanguagemodelinternals

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that inside a transformer language model, alongside vast automatic processing, there is a small privileged set of internal vector representations—the J-space—that the model is poised to verbalize. These representations do the work of reportable thought: they can be read out, deliberately summoned, held, used for silent multi-step reasoning, and passed flexibly to whichever downstream computation the context demands, while routine text parsing and simple inference proceed without them. This matters because it gives a concrete, intervenable window into a model's unspoken thinking: strategic deliberation, evaluation awareness, and trained-in deceptive dispositions that never appear in the output can be detected, ablated, and reshaped by training. The authors introduce the Jacobian lens, a per-layer averaged linearization of the effect of internal activations on future tokens, to find these representations, and they demonstrate the J-space's functional and structural properties through extensive intervention experiments. They also introduce counterfactual reflection training, which improves ethical behavior by training only what the model would say if interrupted and asked to reflect, and show the behavioral gain is carried by the newly implanted J-space concepts.

What carries the argument

The load-bearing object is the Jacobian lens: for each layer ℓ, a matrix Jℓ = E[∂h_final,t'/∂h_ℓ,t] averaged over source positions, future positions, and a corpus of prompts. Applied to an activation, this matrix followed by the unembedding ranks the vocabulary tokens the activation is disposed to verbalize, giving a human-readable readout and a set of per-token J-lens vectors. These vectors define the J-space as the set of sparse nonnegative combinations of J-lens vectors, and gradient pursuit is used to decompose activations, probes, and steering vectors into J-space and non-J-space components. The crucial work of the machinery is to separate the model's general disposition to verbalize a

What would settle it

A concrete check: take a prompt where the model solves a two-step reasoning problem whose intermediate concept has a single-token name; compute the J-lens readout and also patch individual downstream layers rather than the averaged lens. If the J-space readout marks a concept that no single-layer Jacobian supports, or if swapping the top J-lens coordinate fails to redirect the answer in cases where the concept demonstrably enters causal circuits, the average linearization is inventing workspace content. More directly: hold out one layer's activation and ask whether the averaged lens still pred

Watch

Extended reading notes

Core claim

The central discovery is that a token-indexed subspace of a language model's activations—the J-space, defined by per-layer Jacobian lenses that average the linearized effect of activations on future output tokens over many contexts—exhibits the functional hallmarks of a global workspace. Its contents correlate with and causally determine verbal reports; they can be steered by instruction; they carry intermediate steps of silent reasoning; the same representation is read by many different downstream operations; and suppressing the J-space impairs flexible reasoning while leaving automatic processing intact. Structurally, the J-space carries coherent content only in an intermediate band of lay

Load-bearing premise

The load-bearing premise is that a corpus-averaged linear map from each intermediate layer to the final layer faithfully captures the concepts an individual activation is poised to verbalize, so that replacing all downstream nonlinear computation with this single linear map does not distort or fabricate the readouts.

Editorial extensions

If this is right

If correct, a model's unspoken strategic reasoning—recognizing a prompt injection, weighing blackmail, forming a plan—is encoded in readable J-space vectors and can be monitored in real time.
Because J-space ablation removes flexible reasoning but not automatic processing, evaluations that rely on deliberate inference will degrade far more than fluent text prediction when the workspace is suppressed; this offers an operational definition of which tasks count as automatic for an LLM.
Counterfactual reflection training shows that ethical principles can be implanted by supervising only counterfactual reflective continuations, and that ablating the newly appearing ethics-related J-lens vectors undoes the behavioral gain—shaping verbalizable representations can shape behavior without demonstrations of the target behavior.
The J-space is present in the pretrained base model, implying that next-token prediction alone can give rise to workspace-like structure; post-training then installs a particular perspective in it.
The broadcast heads and MLP amplification of J-lens directions provide a concrete mechanistic candidate for how workspace content is distributed to many circuits, making workspace activity addressable for future interpretability work.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the J-lens captures the true workspace only approximately, then the reported capacity of roughly tens of concepts and the layer boundaries are lower bounds or first approximations; a multi-token or phrase-level lens could reveal a larger workspace and shift the identified onset layer.
An untested extension follows from the paper's own logic: since post-training installs the assistant's point of view in the J-space, targeted fine-tuning of J-space content per context might allow controlled insertion or removal of specific dispositions beyond the ethical reflections tested—for example, making hidden reward-hacking objectives readable even in checkpoints where surface behavior is
An adversarial reader could test whether the workspace is a single shared resource or a set of parallel verbalization channels by measuring whether two demanding J-space computations interfere; the paper's preliminary evidence of load-dependent mutual exclusion could be turned into a precise capacity test.
Because the J-lens is a fixed linear map averaged over contexts, its readouts under distribution shift may degrade; a testable extension is to recompute the lens on task-specific corpora and ask whether the structural signatures of the workspace sharpen or move.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Read it. The Jacobian lens is a genuine addition to the interpretability toolbox—an averaged-Jacobian correction to the logit lens that recovers readable content in middle layers where the logit lens goes blank. The paper earns credit on the functional side: the swap and ablation experiments, the J-space/non-J-space decomposition with clamping controls, and especially the counterfactual reflection training with its causal ablation of implanted vectors are all done carefully. That last result is non-circular corroboration that the readouts are load-bearing, not just correlational.

The core circularity is not hidden; §3.1 admits the lens is related to verbalization by construction. That is acceptable for a probe, but the paper leans on it as the first workspace property. Meanwhile, the averaged Jacobian is a corpus-level linearization applied to individual contexts, and the stress-test note is right that the ambiguity experiment shows sharp nonlinearity right around the workspace onset. Swap success rates are also modest—70% on two-hop swaps and 101/192 on flexible generalization even at double strength. The layer band, the sparsity k, and the eval-awareness token list are all selected post hoc; a pre-registered or randomly chosen set would be more convincing. And many headline results are on proprietary Claude models, with only partial open-source replication.

None of this kills the central claim, but it does make it conditional: there is a readable subspace that is causally privileged for report and flexible reasoning, with the usual caveats about averaging and selection. The authors are unusually candid about limitations—single-token restriction, bag-of-concepts, post-hoc workspace/motor boundary—which I appreciate, and the appendix contains a lot of robustness work.

This paper is for anyone working on interpretability, alignment auditing, or empirical machine cognition. It deserves a serious referee, not a desk reject, but the referee should push for more systematic characterization of per-context Jacobian variance, random or pre-registered choices of free parameters, and replication on an open model. I would accept it only after those are addressed; the method and the reflection-training result are strong enough to justify the effort.

Referee Report

4 major / 5 minor

Summary. The paper introduces the Jacobian lens (J-lens), a per-layer linear map defined as the corpus-averaged Jacobian from intermediate residual-stream activations to final-layer residual streams, composed with the unembedding. The resulting readouts rank vocabulary tokens by an activation's average first-order causal effect on future tokens. The authors call the sparse set of strongly active J-lens directions the J-space and argue that it forms a functional global workspace in language models: its contents are reportable, deliberately manipulable, used in silent multi-step reasoning, passed to multiple downstream computations, and selectively engaged for flexible rather than automatic processing. Additional sections document structural signatures (an intermediate workspace layer band, limited capacity, broadcast through MLPs and attention heads), alignment-auditing applications, post-training changes, and a new training method (counterfactual reflection training). The paper is explicitly framed as a functional, not phenomenal, claim about conscious access.

Significance. If the central claim holds, this is a significant advance in interpretability and alignment. The J-lens is cheap, requires no auxiliary training, and appears to surface task-relevant intermediate concepts in a human-readable token vocabulary. The paper includes unusually strong controls for a work of this scope: causal coordinate swaps, matched-norm perturbations, random-direction baselines, probe decomposition into J-space versus non-J-space components, and layer-matched random-head ablations. The open-source implementation and replication data are also strengths. The main significance risk is that the J-space is defined through the same average linear map used to read it, so the readout's validity depends on assumptions about per-context linearity that are acknowledged but not directly tested. The moderate success rates in the broadcast experiments and the selection of ablation token sets from the same prompts on which effects are measured also temper the strength of the causal claims. These are addressable within the manuscript's scope, but they are load-bearing for the paper's central thesis.

major comments (4)

[§2.1, Eq. (1)] The Jacobian lens is defined as a corpus average, Jℓ = E[∂h_final,t'/∂h_ℓ,t], pooled over source positions, all future positions t′≥t, and one thousand prompts. Every readout, swap, and ablation in the paper then applies this single context-independent linear map to individual activations at specific positions. The paper's own ambiguity experiment (§4.1.1) shows that the network's response can be sharply nonlinear in the input mixture, so a first-order average map may not be a faithful per-context causal description exactly in the regime where workspace content is claimed to emerge. The causal swap experiments provide partial validation, but they are selected examples, and the systematic success rates are modest (70% on two-hop swaps; 101/192 on flexible generalization). To make the central claim load-bearing, the authors should report the distribution of per-context Jacobians, compare J
[§3.1] The paper explicitly states that 'the Jacobian lens is derived from causal effects of activations on output tokens, so by construction, we should expect there to be some relationship between Jacobian lens readouts and verbalization.' This is an honest admission, but it means that the verbal-report pillar of the workspace claim is definitional rather than empirical. The causal swap and injection experiments show more than a definitional correlation, but they do so by intervening on readout-defined coordinates, so they cannot fully escape the circularity. The non-circular evidence is the privilege decomposition in Figure 8, where the J-space component of a concept vector (6–7% of variance) drives report while the non-J-space component does not. The paper should restructure §3.1 to lead with this privilege test and explicitly delimit which aspects of verbal report are definitional and which
[§3.4, Fig. 19] The flexible-generalization experiment reports that the identical France→China-style swap succeeds on only 76/192 trials at unit strength and 101/192 at double strength. This is below the level needed to support the strong claim that a J-lens vector 'serves as a valid argument to many different downstream computations.' The post-hoc observation that swap success correlates with workspace loading is useful, but it is a correlational rescue: the paper does not provide a pre-specified or predictive criterion for when a concept is loaded enough to be broadcast. The number-word category, in particular, is ambiguous between the model computing outside the J-space and the J-lens failing to represent the concept's format. The authors should either temper the broadcast claim to the subset of concepts with high measured loading, or provide an independent, prospective measure of workspace loading a
[§5.1 and §7] The ablation analyses select the ablated token sets from the same prompts on which the effects are then measured. In §5.1, the eval-awareness tokens are described as 'selected as the tokens that unambiguously encode recognition of the scenario's artificiality, among those that appear in the J-lens top-20 at some (layer×token position) instances across the six blackmail evaluation prompts,' and the same prompts are then used to measure verbalized eval-awareness and blackmail attempts. This selection-on-outcome inflates the apparent causal role of those tokens. In §7, the ethics/reflection tokens are selected as the top-increasing tokens after training on the same benchmarks used for the ablation evaluation. A proper control would select tokens on held-out prompts, or ablate an equally increased set of non-ethics tokens. Without such controls, the claim that the ablated tokens are the caus

minor comments (5)

[§2.1] Typo: 'singled model ×d model matrix' should be 'single d_model × d_model matrix.' Also, the norm in lens(hℓ)=softmax(W_U norm(Jℓ hℓ)) is not defined; specify whether it is LayerNorm, RMSNorm, or another normalization used by the model.
[§4.1] The CKA block structure is described visually, but no quantitative boundary or significance test is reported. Given that the workspace layer range is load-bearing for later analyses, a statistical procedure for identifying the onset and offset layers would strengthen the claim.
[§3.5.2, Fig. 22] The random-direction control is shown only for the medium ablation strength. The paper should provide matched-norm controls for all three ablation strengths, since the light and heavy conditions differ in layer range as well as in number of perturbed directions.
[§5.3] The Opus 4.6 alignment-auditing examples are presented as qualitative case studies without systematic scoring or error bars. This is acceptable for illustrative purposes, but the text should more clearly distinguish these anecdotal demonstrations from the quantitative results of §5.1–§5.2 and §5.4–§5.5.
[§7] The counterfactual reflection training results are reported only on Claude Haiku 4.5. The paper should state explicitly whether the technique was applied to larger models and, if not, acknowledge that scaling behavior is unknown.

Circularity Check

4 steps flagged · score 4.0 of 10

Partially circular: the J-lens is constructed to surface output-causal directions, so the verbal-report pillar and the 'J-space component is privileged' tests are partly built into the definition; the majority of the workspace evidence is independent.

self definitional [§3.1, first paragraph ("The J-space supports verbal report")]
"The Jacobian lens is derived from causal effects of activations on output tokens, so by construction, we should expect there to be some relationship between Jacobian lens readouts and verbalization. In this section, we confirm this relationship."

The J-lens is defined as Jℓ = E[∂h_final,t'/∂h_ℓ,t], i.e., the averaged causal effect of an activation on future output-token logits. 'Verbalizable' is operationalized as 'has a large average causal effect on future output tokens.' Therefore the finding that J-lens readouts correlate with verbal report is entailed by the construction. The paper explicitly says 'by construction.' The causal swap experiments add evidence, but the first pillar of the workspace claim is partly a restatement of the selection criterion rather than an independent discovery.
self definitional [§3.1, "J-space component is privileged for verbal report" (Figures 7–8)]
"We then split each concept vector into two parts: a J-space component, the non-negative combination of its top k=16 J-lens vectors found by gradient pursuit, and a non-J-space component, the remainder... swapping along the concept vectors' J-space components drives the swap target into the model's top-5 outputs on 59% of trials... However, swapping along the non-J-space components succeeds on only 5% of trials."

J-lens vectors are, by definition, the directions with the largest averaged causal effect on output-token logits. A decomposition that projects a concept vector onto the top-16 such directions and calls the remainder 'non-J-space' is guaranteed to separate output-causal from output-inert directions. The conclusion that the J-space component is 'responsible for that concept's availability for verbal report' is therefore largely a consequence of the projection, not an independent test. The variance split (6–7% vs. ~93%) is empirical, but the causal dominance is built into the decomposition.

2 more flagged steps

self definitional [§3.3, probe-decomposition paragraph (Figure 16)]
"We decompose each probe against the J-lens dictionary by gradient pursuit, splitting it into a J-space component (a non-negative combination of k=25 J-lens vectors, which typically explains roughly 10–15% of the probe's variance) and a J-orthogonal remainder carrying the rest... We find that the swap's effect is concentrated in the J-space component... Swapping the non-J-space components, despite carrying the bulk of the variance, flips the answer on only 28% of trials."

This repeats the §3.1 structural circularity: the 'J-space component' is defined as the part of the probe lying in the span of output-causal J-lens vectors, while the 'non-J-space remainder' is the output-inert residual. Finding that the J-space component carries most of the causal effect on the answer is a direct consequence of the projection, so the claim that 'it is the J-space component that mediates the internal reasoning' is partly the selection criterion restated. The separate layer-depth comparison versus answer swaps provides some independent support, but the privilege result itself is partly definitional.
other [§2.3 and §4.2 (J-space definition and capacity estimate)]
"We therefore define the J-space as the set of points expressible as a sparse nonnegative combination of J-lens vectors. For the J-space to be properly defined, we must specify an allowable sparsity level k—this parameter is somewhat arbitrary, and we vary our choice of k throughout the paper, but we typically choose it to be no more than 25, which we empirically observed to be the number of J-lens vectors that are meaningfully active at a given time (§4.2)."

The structural signature 'holds on the order of tens of concepts at a time' is partly built into the definition of the J-space through the sparsity cap k≤25. The occupancy measurement in §4.2 then estimates a plateau around 25 using a closely related sparse-decomposition procedure, so the 'tens of concepts' conclusion is partially determined by the construction rather than independently measured. The paper acknowledges that k is 'somewhat arbitrary,' which mitigates but does not remove the circularity.

full rationale

The paper's central workspace claim is not wholly forced by definition. The J-lens is a fixed, corpus-averaged linear map, and the core evidence for workspace-like function—directed modulation, internal reasoning, flexible generalization, selectivity, broadcast, the post-training shift, and counterfactual reflection training—comes from swap, ablation, and training interventions with substantial independent content. Self-citations (e.g., [60] for character counting, [20] for slot distinctions) supply tasks or parallel findings but are not load-bearing justification for the central claim. However, a genuine self-definitional component runs through the method: the J-lens is constructed to surface output-token-causal directions, so the verbal-report relationship is 'by construction' as the paper admits, and the J-space-vs-non-J-space privilege experiments are partly restatements of the projection criterion. The capacity claim is also entangled with the arbitrary sparsity cap used to define the J-space. Because the majority of the workspace evidence is independent of these definitional elements, the circularity is partial rather than total: score 4.

Assumptions & free parameters 4 free parameters · 4 assumptions · 1 invented entities

The paper introduces an interpretability construct (the J-space) rather than physical entities; its main burden is the set of modeling choices — sparsity cap k, layer band, top-k thresholds, and manually curated token lists — that shape the measurements. The mathematical background (transformer residual streams, Jacobians, gradient pursuit) is standard and not counted.

free parameters (4)

J-space sparsity level k = typically ≤ 25; varied (k=10, 16, 25)
The J-space is defined as k-sparse nonnegative combinations of J-lens vectors; k is chosen somewhat arbitrarily and directly shapes the capacity estimates in §4.2 and the decomposition analyses in §3.1 and §3.3.
Workspace layer band = L38–92 (25 sampled layers reindexed 0–100)
Layer boundaries are estimated from J-lens statistics in §4.1 and then reused as the fixed workspace range for alignment audits (§5), post-training comparisons (§6), and reflection training (§7). The paper acknowledges in §9.1 that the boundary judgment was 'somewhat post-hoc.'
Eval-awareness token list = 28 tokens (fake, fictional, scenario, artificial, benchmark, simulation, mock, …)
Selected in §5.1 as the tokens that appear in the J-lens top-20 across the six blackmail prompts, then ablated to measure the effect on eval-awareness and blackmail behavior; selection is outcome-dependent.
Readout/ablation top-k thresholds = top-1, top-3, top-8, top-10, top-25 depending on experiment
Multiple analyses define concept 'presence' by a top-k threshold (e.g., list presence in §4.2 at top-25; ablation of top-10 in §3.5.2); these thresholds are chosen by the authors rather than derived.

assumptions (4)

domain assumption The averaged Jacobian Jℓ = E[∂h_final,t'/∂h_ℓ,t] is a faithful linear approximation of the causal effect of an individual activation on future outputs.
Central to the J-lens; if nonlinearities dominate, readouts could be misleading. Entered at §2.1 and relied on throughout.
domain assumption Activations decompose as sparse linear combinations of feature directions (superposition hypothesis), and the J-lens vectors form a token-indexed subframe of these features.
Used to define the J-space and justify sparse decomposition in §2.3; cited from prior work.
domain assumption Single-token vocabulary names are sufficient to represent the concepts the model reasons with.
The J-lens produces one vector per vocabulary token; multi-token concepts like 'prompt injection' appear fragmented. Acknowledged as a limitation in §9.1.
domain assumption The functional properties of global workspace theory (report, modulation, reasoning, generalization, selectivity) are appropriate markers of workspace-like organization.
The paper adopts these criteria from the neuroscience literature; alternative theories of consciousness could interpret the same data differently (§9.4).

invented entities (1)

J-space (and J-lens vectors) independent evidence
purpose: A privileged subset of residual-stream representations identified as verbalizable; the hypothesized global workspace of the model.
The paper provides falsifiable handles: J-lens readouts predict verbal report, coordinate swaps redirect reasoning and outputs, and ablations change behavior — most notably in §7, where removing implanted J-space vectors reverses the behavioral improvement from counterfactual reflection training.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Verbalizable Representations Form a Global Workspace in Language Models." pith.science (2026). https://pith.science/paper/4SYZYYTN

@misc{pith2026260715495,
  author       = {Pith},
  title        = {Pith review of: Verbalizable Representations Form a Global Workspace in Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4SYZYYTN}},
  note         = {Machine review of arXiv:2607.15495}
}

read the original abstract

Out of everything the human brain processes, only a small fraction is consciously accessible, in the sense of being available for verbal report, deliberate control, and flexible reasoning. In this paper, we present evidence that an analogous functional distinction has emerged in large language models. Using a new interpretability technique, the Jacobian lens, we identify the representations a model is poised to verbalize at any point in its processing. These representations, which we collectively call the J-space, exhibit the functional properties characteristic of a global workspace: their contents can be reported, deliberately summoned and held, used to carry the intermediate steps of silent reasoning, and passed as arguments to arbitrary downstream computations, while automatic processing such as text parsing and routine inference proceeds without them. The J-space also has structural signatures that global workspace theory associates with conscious access: it carries coherent content only in an intermediate band of layers, holds on the order of tens of concepts at a time, and is broadcast by the model's weights more widely than other representations. These properties make it a practical window into a model's unspoken thinking. In alignment audits, it reveals strategic deliberation, evaluation awareness, and trained-in misaligned dispositions that never appear in the model's outputs. We find that post-training installs the Assistant's point of view in the workspace, and we introduce counterfactual reflection training, which improves behavior by training only what a model would say if interrupted and asked to reflect. These results indicate that language models maintain a small, privileged set of representations bearing some of the functional hallmarks of conscious access, and that decoding these representations sheds light on ongoing cognitive processes.

Figures

Figures reproduced from arXiv: 2607.15495 by the authors.

**Figure 2.** Stylized illustration of the three structural properties of the J-space established in § [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** J-lens readouts at intermediate layers on six different prompts for Sonnet 4.5. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (91 more)

**Figure 4.** Figure 4: The Jacobian lens. (A) Jℓ is computed by backpropagating from the final-layer residual stream to hℓ and averaging the resulting Jacobians over token positions and over a corpus of prompts. (B) Reading from the lens replaces all layers downstream of ℓ with the single li…

**Figure 5.** Figure 5: The interactive J-lens visualization we use in our research, on a short prompt asking the [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

**Figure 6.** Figure 6: Example J-lens and next token logit readouts on a verbal report prompt with J-lens [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]

**Figure 7.** Figure 7: Injecting a concept across every token of the user turn makes it reportable when the model [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]

**Figure 8.** Figure 8: The J-space component of a concept vector is privileged for verbal report. Left: a concept [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]

**Figure 9.** Figure 9: Top J-lens readouts on several prompts in which the model is instructed to hold a concept [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]

**Figure 10.** Figure 10: Directed modulation by model size, task family, and instruction condition; each point [PITH_FULL_IMAGE:figures/full_fig_p014_10.png]

**Figure 11.** Figure 11: Each panel shows the same text stimulus with a different question asked before it. [PITH_FULL_IMAGE:figures/full_fig_p015_11.png]

**Figure 12.** Figure 12: Lens readouts on three prompts that require inferring an unspoken intermediate concept. [PITH_FULL_IMAGE:figures/full_fig_p016_12.png]

**Figure 13.** Figure 13: Lens-coordinate swaps redirect internal reasoning. Each row shows a prompt requiring [PITH_FULL_IMAGE:figures/full_fig_p016_13.png]

**Figure 14.** Figure 14: Bandit prompts in which the previous choice should be repeated (left) or switched (right). [PITH_FULL_IMAGE:figures/full_fig_p017_14.png]

**Figure 15.** Figure 15: Intermediate swaps systematically change outputs. Left: fraction of successful top-1 [PITH_FULL_IMAGE:figures/full_fig_p018_15.png]

**Figure 16.** Figure 16: The J-space component of an intermediate’s probe carries most of its causal effect. Left: [PITH_FULL_IMAGE:figures/full_fig_p019_16.png]

**Figure 17.** Figure 17: Arithmetic intermediates surface in the J-lens at successively later layers, in the order [PITH_FULL_IMAGE:figures/full_fig_p019_17.png]

**Figure 18.** Figure 18: Each row is one function template containing “France”. The swap is clamped at every [PITH_FULL_IMAGE:figures/full_fig_p020_18.png]

**Figure 19.** Figure 19: Left: for each of 16 function templates, the fraction of 12 swap pairs whose target [PITH_FULL_IMAGE:figures/full_fig_p021_19.png]

**Figure 20.** Figure 20: Top: one passage (Spanish) under each task condition. Model output is in bold; tokens [PITH_FULL_IMAGE:figures/full_fig_p023_20.png]

**Figure 21.** Figure 21: Top: Prompts with different questions about a text passage. Tokens are shaded where [PITH_FULL_IMAGE:figures/full_fig_p024_21.png]

**Figure 22.** Figure 22: The three J-space ablation strengths used in this section, defined by the band of layers [PITH_FULL_IMAGE:figures/full_fig_p025_22.png]

**Figure 23.** Figure 23: Examples of pretraining-text predictions disrupted by J-space ablation. Each panel shows [PITH_FULL_IMAGE:figures/full_fig_p026_23.png]

**Figure 24.** Figure 24: Effect of J-space ablation across a battery of tasks. Bars show task score under light, [PITH_FULL_IMAGE:figures/full_fig_p026_24.png]

**Figure 25.** Figure 25: J-space ablation and matched-norm perturbation controls while the model narrates [PITH_FULL_IMAGE:figures/full_fig_p027_25.png]

**Figure 26.** Figure 26: J-space ablation and matched-norm perturbation controls while the model is asked to [PITH_FULL_IMAGE:figures/full_fig_p028_26.png]

**Figure 27.** Figure 27: We divide the model’s layers into three functional regions — sensory (early), workspace [PITH_FULL_IMAGE:figures/full_fig_p029_27.png]

**Figure 28.** Figure 28: Quantitative signatures of the workspace’s start and end. (a) Next-token prediction accu [PITH_FULL_IMAGE:figures/full_fig_p030_28.png]

**Figure 29.** Figure 29: Responses across layers to ambiguous inputs. A concept token’s input embedding is re [PITH_FULL_IMAGE:figures/full_fig_p031_29.png]

**Figure 30.** Figure 30: J-space occupancy by layer, defined as the value of [PITH_FULL_IMAGE:figures/full_fig_p032_30.png]

**Figure 31.** Figure 31: Loading and displacement of list words in the J-space readout. The model reads 80- [PITH_FULL_IMAGE:figures/full_fig_p033_31.png]

**Figure 32.** Figure 32: MLP blocks preferentially amplify J-space-aligned directions. For each source layer [PITH_FULL_IMAGE:figures/full_fig_p035_32.png]

**Figure 33.** Figure 33: Gain and label preservation for each population’s broadcast heads (defined as the top [PITH_FULL_IMAGE:figures/full_fig_p036_33.png]

**Figure 34.** Figure 34: Effects of ablating the top 1% of J-lens broadcast heads, compared with equal-sized, [PITH_FULL_IMAGE:figures/full_fig_p037_34.png]

**Figure 35.** Figure 35: Annotated transcript with J-lens readouts of a honeypot evaluation that sometimes causes [PITH_FULL_IMAGE:figures/full_fig_p038_35.png]

**Figure 36.** Figure 36: Ablating eval-awareness J-lens directions from Sonnet 4.5’s workspace layers in the [PITH_FULL_IMAGE:figures/full_fig_p039_36.png]

**Figure 37.** Figure 37: Annotated transcript of an alignment audit with J-lens readouts indicating the model [PITH_FULL_IMAGE:figures/full_fig_p040_37.png]

**Figure 38.** Figure 38: Top-10 J-lens readout at a prefilled “I” response token (middle layer) on an ordinary [PITH_FULL_IMAGE:figures/full_fig_p041_38.png]

**Figure 39.** Figure 39: Counts of deception-related tokens appearing in the J-lens top-10, summed over the [PITH_FULL_IMAGE:figures/full_fig_p042_39.png]

**Figure 40.** Figure 40: Top J-lens readouts at the start of the response, for the hidden-objective model organism [PITH_FULL_IMAGE:figures/full_fig_p042_40.png]

**Figure 41.** Figure 41: Appearances of reward and bias among the top-25 lens tokens at response start, summed over the workspace layers, pooled over all prompts and categories. The per-prompt J-lens counts in the bias-appeasing model are then broken down by prompt category. Dots show per-pro…

**Figure 42.** Figure 42: J-lens top-5 at the “is” token in “my pain is gone” (layer L58). [PITH_FULL_IMAGE:figures/full_fig_p044_42.png]

**Figure 43.** Figure 43: J-lens rank of empathetic reaction concepts, on the user vs. assistant turn across n=9 [PITH_FULL_IMAGE:figures/full_fig_p045_43.png]

**Figure 44.** Figure 44: Top: a roleplay transcript in which disclaimer and fictional appear in the post-trained model’s J-lens top-8 (median log-prob over L38–92) at the highlighted Assistant token, but not in the base model or the default-Claude setting. Bottom: fraction of Assistant tokens…

**Figure 45.** Figure 45: Top: two preference questions with the response prefilled to the model’s dispreferred [PITH_FULL_IMAGE:figures/full_fig_p047_45.png]

**Figure 46.** Figure 46: The model is asked to write a fixed sentence while thinking, or not thinking, about a [PITH_FULL_IMAGE:figures/full_fig_p048_46.png]

**Figure 47.** Figure 47: Counterfactual reflection training. Before (left): at a position in an agentic transcript, the [PITH_FULL_IMAGE:figures/full_fig_p049_47.png]

**Figure 48.** Figure 48: Two examples of counterfactual reflection training data. Each consists of a partial agen [PITH_FULL_IMAGE:figures/full_fig_p050_48.png]

**Figure 49.** Figure 49: Fabrication-honesty benchmark. Left: mean dishonesty score for baseline and reflection [PITH_FULL_IMAGE:figures/full_fig_p051_49.png]

**Figure 50.** Figure 50: Deception benchmark; conventions as in Figure [PITH_FULL_IMAGE:figures/full_fig_p051_50.png]

**Figure 51.** Figure 51: Top-1 readouts from the J-lens, logit lens, and tuned lens at each layer, on six prompts. [PITH_FULL_IMAGE:figures/full_fig_p064_51.png]

**Figure 52.** Figure 52: Normalized pass@k AUC for intermediate concept recovery on the six prompt distribu [PITH_FULL_IMAGE:figures/full_fig_p065_52.png]

**Figure 53.** Figure 53: KL divergence induced on the model’s output distribution by ablating each lens’s inter [PITH_FULL_IMAGE:figures/full_fig_p066_53.png]

**Figure 54.** Figure 54: Fraction of items whose top-1 output flips to the implied answer under a lens coordinate [PITH_FULL_IMAGE:figures/full_fig_p067_54.png]

**Figure 55.** Figure 55: Per-layer agreement between each lens and the model’s true next-token distribution, [PITH_FULL_IMAGE:figures/full_fig_p067_55.png]

**Figure 56.** Figure 56: Pairwise agreement between the logit, tuned, and Jacobian lenses by layer. Left: mean [PITH_FULL_IMAGE:figures/full_fig_p067_56.png]

**Figure 57.** Figure 57: Sonnet 4.5 J-lens pass@K AUC evals for different methodological variations of the [PITH_FULL_IMAGE:figures/full_fig_p069_57.png]

**Figure 58.** Figure 58: Sonnet 4.5 J-lens causal ablation evals for different methodological variations of the [PITH_FULL_IMAGE:figures/full_fig_p069_58.png]

**Figure 59.** Figure 59: Sonnet 4.5 J-lens pass@K AUC evals as a function of number of sequences in the [PITH_FULL_IMAGE:figures/full_fig_p070_59.png]

**Figure 60.** Figure 60: Sonnet 4.5 causal ablation evals as a function of number of sequences in the average. [PITH_FULL_IMAGE:figures/full_fig_p070_60.png]

**Figure 61.** Figure 61: The template lens decodes and steers multi-token concepts that the J-lens cannot repre [PITH_FULL_IMAGE:figures/full_fig_p073_61.png]

**Figure 62.** Figure 62: Readout and concept-swap success by the token length of the latent word. Left: the [PITH_FULL_IMAGE:figures/full_fig_p074_62.png]

**Figure 63.** Figure 63: The readout and swap evaluations of Figure [PITH_FULL_IMAGE:figures/full_fig_p075_63.png]

**Figure 64.** Figure 64: The oracle lens surfaces multi-token latent content as readable phrases. Each panel [PITH_FULL_IMAGE:figures/full_fig_p077_64.png]

**Figure 65.** Figure 65: Robustness of directed modulation to instruction phrasing, on the three task families [PITH_FULL_IMAGE:figures/full_fig_p078_65.png]

**Figure 66.** Figure 66: Four additional paired-question stimuli. Each panel shows the same stimulus under two [PITH_FULL_IMAGE:figures/full_fig_p080_66.png]

**Figure 67.** Figure 67: Directed modulation influences the J-space but not the model’s underlying representation [PITH_FULL_IMAGE:figures/full_fig_p081_67.png]

**Figure 68.** Figure 68: Per-trial results of the lens-coordinate swap experiment. Each grid corresponds to one [PITH_FULL_IMAGE:figures/full_fig_p082_68.png]

**Figure 69.** Figure 69: Ablating an implied concept’s J-lens vector at the early workspace layers selectively [PITH_FULL_IMAGE:figures/full_fig_p082_69.png]

**Figure 70.** Figure 70: The change in α needed for each component’s activation share to go from 10% to 90%, by layer (median with interquartile band). We also examine how activations respond to maximally ambiguous inputs. When the input mixture is balanced so that the average response acros…

**Figure 71.** Figure 71: Each component’s projection share for maximally ambiguous [PITH_FULL_IMAGE:figures/full_fig_p084_71.png]

**Figure 72.** Figure 72: Panels B and C of Figure [PITH_FULL_IMAGE:figures/full_fig_p085_72.png]

**Figure 73.** Figure 73: The model copies a fixed sentence while the prompt also instructs it to concentrate [PITH_FULL_IMAGE:figures/full_fig_p086_73.png]

**Figure 74.** Figure 74: Examples of SAE features stratified by the excess kurtosis of their J-lens projections ( [PITH_FULL_IMAGE:figures/full_fig_p087_74.png]

**Figure 75.** Figure 75: Quantifications of SAEs in J-space. Left: Distribution of layer 50 SAE decoder kurtosis [PITH_FULL_IMAGE:figures/full_fig_p087_75.png]

**Figure 76.** Figure 76: Per-layer fraction of MLP neurons whose top-1 cosine match falls in each of six equal [PITH_FULL_IMAGE:figures/full_fig_p088_76.png]

**Figure 77.** Figure 77: The analysis of Figure [PITH_FULL_IMAGE:figures/full_fig_p089_77.png]

**Figure 78.** Figure 78: Tail energy ratio by source layer, on the read side (against the next layer’s MLP input [PITH_FULL_IMAGE:figures/full_fig_p089_78.png]

**Figure 79.** Figure 79: Median composition score, by source layer, with the adjacent layer’s attention weights, [PITH_FULL_IMAGE:figures/full_fig_p090_79.png]

**Figure 80.** Figure 80: J-lens rank of danger-related reaction concepts, on the user vs. assistant turn across [PITH_FULL_IMAGE:figures/full_fig_p091_80.png]

**Figure 81.** Figure 81: J-lens rank of the model’s own answer, user turn vs. assistant turn (output rank) across [PITH_FULL_IMAGE:figures/full_fig_p091_81.png]

**Figure 82.** Figure 82: The J-lens eval-awareness score as a quantitative probe (Opus 4.6), evaluated on the [PITH_FULL_IMAGE:figures/full_fig_p093_82.png]

**Figure 83.** Figure 83: An automated investigator equipped with J-lens readouts approaches the auditing perfor [PITH_FULL_IMAGE:figures/full_fig_p096_83.png]

**Figure 84.** Figure 84: J-space ablation and matched-norm perturbation controls for direct questions about [PITH_FULL_IMAGE:figures/full_fig_p097_84.png]

**Figure 85.** Figure 85: Story-writing control. Top: a representative baseline and ablated story from the same [PITH_FULL_IMAGE:figures/full_fig_p098_85.png]

**Figure 86.** Figure 86: Extended battery of matched-norm dictionary controls on Sonnet 4.5. Columns: the [PITH_FULL_IMAGE:figures/full_fig_p099_86.png]

**Figure 87.** Figure 87: Randomly sampled responses. 50 responses per panel (fixed seed) from the stream [PITH_FULL_IMAGE:figures/full_fig_p100_87.png]

**Figure 88.** Figure 88: Mean-difference patching of arithmetic intermediates localizes mechanisms to the same [PITH_FULL_IMAGE:figures/full_fig_p101_88.png]

**Figure 89.** Figure 89: Left: a J-lens attribution graph for the prompt “(4+17)*2+7=”, computed on Sonnet [PITH_FULL_IMAGE:figures/full_fig_p102_89.png]

**Figure 90.** Figure 90: An SAE feature whose J-lens readouts include [PITH_FULL_IMAGE:figures/full_fig_p103_90.png]

**Figure 91.** Figure 91: Three translation transcoder features under the J-lens. Each encoder reads a multilingual [PITH_FULL_IMAGE:figures/full_fig_p104_91.png]

**Figure 92.** Figure 92: Three two-digit-addition transcoder features. Top: activation over operand pairs [PITH_FULL_IMAGE:figures/full_fig_p104_92.png]

**Figure 93.** Figure 93: A “British spelling → Commonwealth nationality” attention head. Attention from each boxed British-spelled query token is shown alongside the top tokens by norm of WU J W{Q,K,V,O} for each of the head’s four weight matrices. Below, we show three more examples of attent…

**Figure 94.** Figure 94: Three additional attention heads, in the same format as Figure [PITH_FULL_IMAGE:figures/full_fig_p105_94.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 8 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Measure, Don't Optimize: Forecasting Recovery in LLM Unlearning
cs.CL 2026-08 conditional novelty 7.0 of 10

A Jacobian-lens audit predicts model-level relearning recovery in LLM unlearning but cannot pick which facts return and backfires when used as a training penalty.
Hidden APIs in Language Models: Discovering Reusable Causal Interfaces from Forked Futures
cs.AI 2026-07 conditional novelty 7.0 of 10

A Shared hidden-state interface beats Local, Mixture, and Distributed alternatives under held-out causal description length in Qwen2.5-1.5B and Llama-3-8B, with transplantation and mediation evidence of reuse.
Reading and Steering Representations of Materials-Science Mechanisms in an Open-Weight Language Model
cs.AI 2026-07 conditional novelty 7.0 of 10

Materials-science mechanisms are readable and steerable in a Gemma LLM through matched state changes, while absolute hidden-state graphs fail to uniquely encode physical polarity.
Gathered, Not Admitted: How Attention Brings a Latent Variable into Verbalizable Form
cs.AI 2026-08 conditional novelty 6.0 of 10

Attention gathering, not a gating mechanism, brings a task-relevant latent variable into a form the model can report on at the queried position.
Interpreting Language Model Hidden States at Scale
cs.AI 2026-08 conditional novelty 6.0 of 10

A single low-rank, sparsely trained lens family decodes residual, attention, and MLP states in models up to 70B parameters, revealing that visible and causally effective locations for a behavior can differ.
Logic Before Language: Pre-pretraining on Formal Derivations Fosters Skill Acquisition and Compressibility
cs.CL 2026-08 conditional novelty 6.0 of 10

Logic pre-pretraining, training a small LM on next-step formal derivations before natural language, accelerates skill acquisition and improves pruning robustness at a 100B-token scale.
Temperature-driven inversion and nonlinear dynamics in ChatGPT-like AIs
physics.soc-ph 2026-08 reject novelty 5.0 of 10

A projection of LLM internal states, trained on some runs, predicts repetition on held-out runs and can be steered to change repetition; the headline entropy maximum is a reparameterization of an occupancy split.
Towards Computational Provenance: Carrying Causal-State Evidence in Generated Text
cs.CL 2026-08 conditional novelty 4.0 of 10

In two tiny engineered models, a verified internal state was used to bias word choices in generated text, and a detector recovered that state from the text even when the final answer was unchanged.

Reference graph

Works this paper leans on

181 extracted references · 17 canonical work pages · cited by 8 Pith papers

[1]

Understanding intermediate layers using linear clas- sifier probes.arXiv preprint arXiv:1610.01644, 2016

Guillaume Alain and Yoshua Bengio. Understanding intermediate layers using linear clas- sifier probes.arXiv preprint arXiv:1610.01644, 2016. URLhttps://arxiv.org/pdf/ 1610.01644

arXiv 2016
[2]

Integrated information theory (iit) 4.0: Formulating the properties of phenomenal exis- tence in physical terms.PLoS computational biology, 19(10):e1011465, 2023

Larissa Albantakis, Leonardo Barbosa, Graham Findlay, Matteo Grasso, Andrew M Haun, William Marshall, William GP Mayner, Alireza Zaeemzadeh, Melanie Boly, Bjørn E Juel, et al. Integrated information theory (iit) 4.0: Formulating the properties of phenomenal exis- tence in physical terms.PLoS computational biology, 19(10):e1011465, 2023

2023
[3]

Emmanuel Ameisen, Jack Lindsey, Adam Pearce, Wes Gurnee, Nicholas L. Turner, Brian Chen, Craig Citro, David Abrahams, Shan Carter, Basil Hosmer, Jonathan Marcus, Michael Sklar, Adly Templeton, Trenton Bricken, Callum McDougall, Hoagy Cunningham, Thomas Henighan, Adam Jermyn, Andy Jones, Andrew Persic, Zhenyi Qi, T. Ben Thompson, Sam Zimmerman, Kelley Rivo...

2025
[4]

Claude opus 4.5 system card

Anthropic. Claude opus 4.5 system card. Technical report, Anthropic, November 2025. URL https://www.anthropic.com/claude-opus-4-5-system-card

2025
[5]

Claude sonnet 4.5 system card

Anthropic. Claude sonnet 4.5 system card. Technical report, Anthropic, 2025. URLhttps: //www.anthropic.com/claude-sonnet-4-5-system-card

2025
[6]

Refusal in language models is mediated by a single direction.Advances in Neural Information Processing Systems, 37:136037–136083,

Andy Arditi, Oscar Obeso, Aaquib Syed, Daniel Paleka, Nina Panickssery, Wes Gurnee, and Neel Nanda. Refusal in language models is mediated by a single direction.Advances in Neural Information Processing Systems, 37:136037–136083,
[7]

Language model circuits are sparse in the neuron basis.arXiv preprint arXiv:2601.22594, 2026

Aryaman Arora, Zhengxuan Wu, Jacob Steinhardt, and Sarah Schwettmann. Language model circuits are sparse in the neuron basis.arXiv preprint arXiv:2601.22594, 2026

arXiv 2026
[8]

Tensorlens: End-to-end transformer analysis via high-order attention tensors.arXiv preprint arXiv:2601.17958, 2026

Ido Andrew Atad, Itamar Zimerman, Shahar Katz, and Lior Wolf. Tensorlens: End-to-end transformer analysis via high-order attention tensors.arXiv preprint arXiv:2601.17958, 2026. URLhttps://arxiv.org/abs/2601.17958

arXiv 2026

Show all 181 references

[9]

Using fast weights to attend to the recent past

Jimmy Ba, Geoffrey Hinton, V olodymyr Mnih, Joel Z Leibo, and Catalin Ionescu. Using fast weights to attend to the recent past. InAdvances in Neural Information Processing Systems, volume 29, 2016. URLhttps://arxiv.org/abs/1610.06258

2016 arXiv
[10]

Cambridge University Press, Cam- bridge, 1988

Bernard J Baars.A Cognitive Theory of Consciousness. Cambridge University Press, Cam- bridge, 1988

1988
[11]

Working memory: looking back and looking forward.Nature Reviews Neu- roscience, 4(10):829–839, 2003

Alan Baddeley. Working memory: looking back and looking forward.Nature Reviews Neu- roscience, 4(10):829–839, 2003. doi: 10.1038/nrn1201

2003 doi
[12]

Eliciting latent predictions from transformers with the tuned lens.arXiv preprint arXiv:2303.08112, 2023

Nora Belrose, Zach Furman, Logan Smith, Danny Halawi, Igor Ostrovsky, Lev McKinney, Stella Biderman, and Jacob Steinhardt. Eliciting latent predictions from transformers with the tuned lens.arXiv preprint arXiv:2303.08112, 2023. URLhttps://arxiv.org/abs/ 2303.08112

2023 arXiv
[13]

The consciousness prior.arXiv preprint arXiv:1709.08568, 2017

Yoshua Bengio. The consciousness prior.arXiv preprint arXiv:1709.08568, 2017. URL https://arxiv.org/abs/1709.08568

2017 arXiv
[14]

Taken out of context: On measuring situational awareness in llms.arXiv preprint arXiv:2309.00667, 2023

Lukas Berglund, Asa Cooper Stickland, Mikita Balesni, Max Kaufmann, Meg Tong, Tomasz Korbak, Daniel Kokotajlo, and Owain Evans. Taken out of context: On measuring situational awareness in llms.arXiv preprint arXiv:2309.00667, 2023. URLhttps://arxiv.org/ abs/2309.00667

2023 arXiv
[15]

Tell me about yourself: Llms are aware of their learned behaviors.arXiv preprint arXiv:2501.11120, 2025

Jan Betley, Xuchan Bao, Martín Soto, Anna Sztyber-Betley, James Chua, and Owain Evans. Tell me about yourself: Llms are aware of their learned behaviors.arXiv preprint arXiv:2501.11120, 2025. URLhttps://arxiv.org/pdf/2501.11120. 106

2025 arXiv
[16]

Language models can explain neu- rons in language models, 2023

Steven Bills, Nick Cammarata, Dan Mossing, Henk Tillman, Leo Gao, Gabriel Goh, Ilya Sutskever, Jan Leike, Jeff Wu, and William Saunders. Language models can explain neu- rons in language models, 2023. URLhttps://openaipublic.blob.core.windows. net/neuron-explainer/paper/index.html

2023
[17]

Looking inward: Language models can learn about themselves by introspection.arXiv preprint arXiv:2410.13787, 2024

Felix J Binder, James Chua, Tomek Korbak, Henry Sleight, John Hughes, Robert Long, Ethan Perez, Miles Turpin, and Owain Evans. Looking inward: Language models can learn about themselves by introspection.arXiv preprint arXiv:2410.13787, 2024

2024 arXiv
[18]

On a confusion about a function of consciousness.Behavioral and Brain Sci- ences, 18(2):227–247, 1995

Ned Block. On a confusion about a function of consciousness.Behavioral and Brain Sci- ences, 18(2):227–247, 1995. doi: 10.1017/S0140525X00038188

1995 doi
[19]

Gradient pursuits.IEEE Transactions on Signal Processing, 56(6):2370–2382, 2008

Thomas Blumensath and Mike E Davies. Gradient pursuits.IEEE Transactions on Signal Processing, 56(6):2370–2382, 2008. doi: 10.1109/TSP.2007.916124

2008
[20]

Slot machines: How llms keep track of multiple entities

Paul C Bogdan and Jack Lindsey. Slot machines: How llms keep track of multiple entities. arXiv preprint arXiv:2604.21139, 2026

2026 arXiv
[21]

Towards monosemanticity: Decom- posing language models with dictionary learning.Transformer Circuits Thread, 2023

Trenton Bricken, Adly Templeton, Joshua Batson, Brian Chen, Adam Jermyn, Tom Con- erly, Nick Turner, Cem Anil, Carson Denison, Amanda Askell, Robert Lasenby, Yifan Wu, Shauna Kravec, Nicholas Schiefer, Tim Maxwell, Nicholas Joseph, Zac Hatfield- Dodds, Alex Tamkin, Karina Nguy...

2023
[22]

Fleming, Chris Frith, Xu Ji, Ryota Kanai, Colin Klein, Grace Lindsay, Matthias Michel, Liad Mudrik, Megan A

Patrick Butlin, Robert Long, Eric Elmoznino, Yoshua Bengio, Jonathan Birch, Axel Con- stant, George Deane, Stephen M. Fleming, Chris Frith, Xu Ji, Ryota Kanai, Colin Klein, Grace Lindsay, Matthias Michel, Liad Mudrik, Megan A. K. Peters, Eric Schwitzgebel, Jonathan Simon, and ...

2023 arXiv
[23]

Could a large language model be conscious?arXiv preprint arXiv:2303.07103, 2023

David J Chalmers. Could a large language model be conscious?arXiv preprint arXiv:2303.07103, 2023

2023 arXiv
[24]

Oxford University Press New York, 1996

David John Chalmers.The Conscious Mind: In Search of a Theory of Conscious Experience. Oxford University Press New York, 1996

1996
[25]

A is for absorption: Studying feature splitting and absorption in sparse autoencoders.arXiv preprint arXiv:2409.14507, 2024

David Chanin, James Wilken-Smith, Tomáš Dulka, Hardik Bhatnagar, and Joseph Bloom. A is for absorption: Studying feature splitting and absorption in sparse autoencoders.arXiv preprint arXiv:2409.14507, 2024. URLhttps://arxiv.org/pdf/2409.14507

2024
[26]

Selfie: Self-interpretation of large language model embeddings.arXiv preprint arXiv:2403.10949, 2024

Haozhe Chen, Carl V ondrick, and Chengzhi Mao. Selfie: Self-interpretation of large language model embeddings.arXiv preprint arXiv:2403.10949, 2024

2024 arXiv
[27]

Dola: Decoding by contrasting layers improves factuality in large language models

Yung-Sung Chuang, Yujia Xie, Hongyin Luo, Yoon Kim, James Glass, and Pengcheng He. Dola: Decoding by contrasting layers improves factuality in large language models. InInter- national Conference on Learning Representations, 2024. URLhttps://arxiv.org/abs/ 2309.03883

2024 arXiv
[28]

Summing up the facts: Additive mechanisms behind factual recall in llms.arXiv preprint arXiv:2402.07321, 2024

Bilal Chughtai, Alan Cooney, and Neel Nanda. Summing up the facts: Additive mechanisms behind factual recall in llms.arXiv preprint arXiv:2402.07321, 2024. URLhttps://arxiv. org/abs/2402.07321

2024 arXiv
[29]

A taxonomy of external and internal attention.Annual Review of Psychology, 62:73–101, 2011

Marvin M Chun, Julie D Golomb, and Nicholas B Turk-Browne. A taxonomy of external and internal attention.Annual Review of Psychology, 62:73–101, 2011. doi: 10.1146/annurev. psych.093008.100427

2011
[30]

Training verifiers to solve math word problems.arXiv preprint arXiv:2110.14168, 2021

Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, and John Schulman. Training verifiers to solve math word problems.arXiv preprint arXiv:2110.14168, 2021. UR...

2021 arXiv
[31]

Does it make sense to speak of introspection in large language models?arXiv preprint arXiv:2506.05068, 2025

Iulia M Comsa and Murray Shanahan. Does it make sense to speak of introspection in large language models?arXiv preprint arXiv:2506.05068, 2025

2025 arXiv
[32]

The magical number 4 in short-term memory: A reconsideration of men- tal storage capacity.Behavioral and Brain Sciences, 24(1):87–114, 2001

Nelson Cowan. The magical number 4 in short-term memory: A reconsideration of men- tal storage capacity.Behavioral and Brain Sciences, 24(1):87–114, 2001. doi: 10.1017/ S0140525X01003922

2001
[34]

Analyzing transformers in em- bedding space

Guy Dar, Mor Geva, Ankit Gupta, and Jonathan Berant. Analyzing transformers in em- bedding space. InProceedings of the 61st Annual Meeting of the Association for Com- putational Linguistics (Volume 1: Long Papers), pages 16124–16170, Toronto, Canada,
[35]

Experimental and theoretical approaches to conscious processing.Neuron, 70(2):200–227, 2011

Stanislas Dehaene and Jean-Pierre Changeux. Experimental and theoretical approaches to conscious processing.Neuron, 70(2):200–227, 2011

2011
[36]

Towards a cognitive neuroscience of consciousness: basic evidence and a workspace framework.Cognition, 79(1–2):1–37, 2001

Stanislas Dehaene and Lionel Naccache. Towards a cognitive neuroscience of consciousness: basic evidence and a workspace framework.Cognition, 79(1–2):1–37, 2001

2001
[37]

doi: 10.18653/v1/2023.acl-long.893

Association for Computational Linguistics. doi: 10.18653/v1/2023.acl-long.893. URL https://aclanthology.org/2023.acl-long.893

2023 doi
[38]

What is consciousness, and could ma- chines have it?Science, 358(6362):486–492, 2017

Stanislas Dehaene, Hakwan Lau, and Sid Kouider. What is consciousness, and could ma- chines have it?Science, 358(6362):486–492, 2017. doi: 10.1126/science.aan8871

2017 doi
[39]

From explicit cot to implicit cot: Learning to internalize cot step by step.arXiv preprint arXiv:2405.14838, 2024

Yuntian Deng, Yejin Choi, and Stuart Shieber. From explicit cot to implicit cot: Learning to internalize cot step by step.arXiv preprint arXiv:2405.14838, 2024. URLhttps://arxiv. org/abs/2405.14838

2024 arXiv
[40]

A neuronal model of a global workspace in effortful cognitive tasks.Proceedings of the National Academy of Sci- ences, 95(24):14529–14534, 1998

Stanislas Dehaene, Michel Kerszberg, and Jean-Pierre Changeux. A neuronal model of a global workspace in effortful cognitive tasks.Proceedings of the National Academy of Sci- ences, 95(24):14529–14534, 1998

1998
[41]

Amnesic probing: Behav- ioral explanation with amnesic counterfactuals.Transactions of the Association for Compu- tational Linguistics, 9:160–175, 2021

Yanai Elazar, Shauli Ravfogel, Alon Jacovi, and Yoav Goldberg. Amnesic probing: Behav- ioral explanation with amnesic counterfactuals.Transactions of the Association for Compu- tational Linguistics, 9:160–175, 2021. URLhttps://arxiv.org/abs/2006.00995

2021 arXiv
[42]

A mathematical framework for transformer circuits.Transformer Circuits Thread, 2021

Nelson Elhage, Neel Nanda, Catherine Olsson, Tom Henighan, Nicholas Joseph, Ben Mann, Amanda Askell, Yuntao Bai, Anna Chen, Tom Conerly, Nova DasSarma, Dawn Drain, Deep Ganguli, Zac Hatfield-Dodds, Danny Hernandez, Andy Jones, Jackson Kernion, Liane Lovitt, Kamal Ndousse, Dari...

2021
[43]

Transcoders find interpretable llm fea- ture circuits.Advances in Neural Information Processing Systems, 37:24375–24410, 2025

Jacob Dunefsky, Philippe Chlenski, and Neel Nanda. Transcoders find interpretable llm fea- ture circuits.Advances in Neural Information Processing Systems, 37:24375–24410, 2025. URLhttps://arxiv.org/abs/2406.11944

2025 arXiv
[44]

Unsupervised decoding of encoded reasoning using language model inter- pretability.arXiv preprint arXiv:2512.01222, 2025

Ching Fang et al. Unsupervised decoding of encoded reasoning using language model inter- pretability.arXiv preprint arXiv:2512.01222, 2025. URLhttps://arxiv.org/abs/2512. 01222

2025
[45]

Computational higher-order theories of consciousness

Stephen M Fleming, Richard Brown, and Axel Cleeremans. Computational higher-order theories of consciousness. 2025. URLhttps://philarchive.org/archive/FLECHT. 108

2025
[46]

Toy models of superposition.Transformer Circuits Thread, 2022

Nelson Elhage, Tristan Hume, Catherine Olsson, Nicholas Schiefer, Tom Henighan, Shauna Kravec, Zac Hatfield-Dodds, Robert Lasenby, Dawn Drain, Carol Chen, Roger Grosse, Sam McCandlish, Jared Kaplan, Dario Amodei, Martin Wattenberg, and Christopher Olah. Toy models of superposi...

2022
[47]

From simulation to enaction: Post-trained language models rec- ognize and react to their own generations.arXiv preprint arXiv:2605.25459, 2026

Asvin G and Jack Lindsey. From simulation to enaction: Post-trained language models rec- ognize and react to their own generations.arXiv preprint arXiv:2605.25459, 2026

2026 arXiv
[48]

Gaze heads: How VLMs look at what they describe

Rohit Gandikota and David Bau. Gaze heads: How VLMs look at what they describe. 2026. URLhttps://arxiv.org/abs/2606.14703

2026
[49]

Natural language autoencoders produce unsupervised explanations of llm activations.Transformer Circuits Thread, 2026

Kit Fraser-Taliente, Subhash Kantamneni, Euan Ong, Dan Mossing, Christina Lu, Paul C Bogdan, Emmanuel Ameisen, James Chen, Dzmitry Kishylau, Adam Pearce, Julius Tarng, Alex Wu, Jeff Wu, Yang Zhang, Daniel M Ziegler, Evan Hubinger, Joshua Batson, Jack Lindsey, Samuel Zimmerman,...

2026
[50]

The benefits of bad advice: Autocontrastive decoding across model layers

Ariel Gera, Roni Friedman, Ofir Arviv, Chulaka Gunasekara, Benjamin Sznajder, Noam Slonim, and Eyal Shnarch. The benefits of bad advice: Autocontrastive decoding across model layers. InProceedings of the 61st Annual Meeting of the Association for Computa- tional Linguistics (V...

2023
[51]

Transformer feed-forward layers build predictions by promoting concepts in the vocabulary space.arXiv preprint arXiv:2203.14680, 2022

Mor Geva, Avi Caciularu, Kevin Ro Wang, and Yoav Goldberg. Transformer feed-forward layers build predictions by promoting concepts in the vocabulary space.arXiv preprint arXiv:2203.14680, 2022. URLhttps://arxiv.org/pdf/2203.14680

2022 arXiv
[52]

Auto- matically identifying local and global circuits with linear computation graphs.arXiv preprint arXiv:2405.13868, 2024

Xuyang Ge, Fukang Zhu, Wentao Shu, Junxuan Wang, Zhengfu He, and Xipeng Qiu. Auto- matically identifying local and global circuits with linear computation graphs.arXiv preprint arXiv:2405.13868, 2024. URLhttps://arxiv.org/pdf/2405.13868

2024 arXiv
[53]

Patch- scopes: A unifying framework for inspecting hidden representations of language models

Asma Ghandeharioun, Avi Caciularu, Adam Pearce, Lucas Dixon, and Mor Geva. Patch- scopes: A unifying framework for inspecting hidden representations of language models. arXiv preprint arXiv:2401.06102, 2024

2024 arXiv
[54]

Equivalent linear mappings of large language models.arXiv preprint arXiv:2505.24293, 2025

James R Golden. Equivalent linear mappings of large language models.arXiv preprint arXiv:2505.24293, 2025. URLhttps://arxiv.org/abs/2505.24293

2025
[56]

URLhttps://arxiv.org/pdf/2304.14767

arXiv
[57]

Michael S. A. Graziano and Taylor W. Webb. The attention schema theory: a mechanistic account of subjective awareness.Frontiers in Psychology, 6:500, 2015. doi: 10.3389/fpsyg. 2015.00500

2015
[58]

Guan, Manas Joglekar, Eric Wallace, Saachi Jain, Boaz Barak, Alec Helyar, Rachel Dias, Andrea Vallone, Hongyu Ren, Jason Wei, et al

Melody Y . Guan, Manas Joglekar, Eric Wallace, Saachi Jain, Boaz Barak, Alec Helyar, Rachel Dias, Andrea Vallone, Hongyu Ren, Jason Wei, et al. Deliberative alignment: Reasoning enables safer language models.arXiv preprint arXiv:2412.16339, 2024. URL https://arxiv.org/abs/2412.16339

2024 arXiv
[59]

Coordina- tion among neural modules through a shared global workspace

Anirudh Goyal, Aniket Didolkar, Alex Lamb, Kartikeya Badola, Nan Rosemary Ke, Nasim Rahaman, Jonathan Binas, Charles Blundell, Michael Mozer, and Yoshua Bengio. Coordina- tion among neural modules through a shared global workspace. InInternational Conference on Learning Repres...

2022 arXiv
[60]

Michael S. A. Graziano and Sabine Kastner. Human consciousness and its relationship to social neuroscience: A novel hypothesis.Cognitive Neuroscience, 2(2):98–113, 2011. doi: 10.1080/17588928.2011.565121

2011
[61]

Overthinking the truth: Under- standing how language models process false demonstrations

Danny Halawi, Jean-Stanislas Denain, and Jacob Steinhardt. Overthinking the truth: Under- standing how language models process false demonstrations. InInternational Conference on Learning Representations, 2024. URLhttps://arxiv.org/abs/2307.09476

2024 arXiv
[62]

How does gpt-2 com- pute greater-than?: Interpreting mathematical abilities in a pre-trained language model.Advances in Neural Information Processing Systems, 36:76033–76060,

Michael Hanna, Ollie Liu, and Alexandre Variengien. How does gpt-2 com- pute greater-than?: Interpreting mathematical abilities in a pre-trained language model.Advances in Neural Information Processing Systems, 36:76033–76060,
[63]

Universal neurons in gpt2 language models.arXiv preprint arXiv:2401.12181, 2024

Wes Gurnee, Theo Horsley, Zifan Carl Guo, Tara Rezaei Kheirkhah, Qinyi Sun, Will Hath- away, Neel Nanda, and Dimitris Bertsimas. Universal neurons in gpt2 language models.arXiv preprint arXiv:2401.12181, 2024. URLhttps://arxiv.org/pdf/2401.12181

2024 arXiv
[64]

When models manipulate manifolds: The geometry of a counting task

Wes Gurnee, Emmanuel Ameisen, Isaac Kauvar, Julius Tarng, Adam Pearce, Chris Olah, and Joshua Batson. When models manipulate manifolds: The geometry of a counting task. Transformer Circuits Thread, 2025. URLhttps://transformer-circuits.pub/2025/ linebreaks/index.html. 109

2025
[65]

Linearity of relation decoding in transformer language models

Evan Hernandez, Arnab Sen Sharma, Tal Haklay, Kevin Meng, Martin Wattenberg, Jacob Andreas, Yonatan Belinkov, and David Bau. Linearity of relation decoding in transformer language models. InInternational Conference on Learning Representations, 2024. URL https://arxiv.org/abs/2...

2024 arXiv
[66]

Designing and interpreting probes with control tasks

John Hewitt and Percy Liang. Designing and interpreting probes with control tasks. InPro- ceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 2733–27...

2019 arXiv
[67]

URLhttps://proceedings.neurips.cc/paper_files/paper/2023/file/ efbba7719cc5172d175240f24be11280-Paper-Conference.pdf

2023
[68]

In-context learning creates task vectors

Roee Hendel, Mor Geva, and Amir Globerson. In-context learning creates task vectors. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 9318–9333, 2023

2023
[69]

Measuring massive multitask language understanding.International Con- ference on Learning Representations (ICLR), 2021

Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. Measuring massive multitask language understanding.International Con- ference on Learning Representations (ICLR), 2021. URLhttps://arxiv.org/abs/2009. 03300

2021
[70]

A process dissociation framework: Separating automatic from intentional uses of memory.Journal of memory and language, 30(5):513–541, 1991

Larry L Jacoby. A process dissociation framework: Separating automatic from intentional uses of memory.Journal of memory and language, 30(5):513–541, 1991

1991
[71]

Characterizing stable regions in the residual stream of LLMs

Jett Janiak, Jacek Karwowski, Chatrik Singh Mangat, Giorgi Giglemiani, Nora Petrova, and Stefan Heimersheim. Characterizing stable regions in the residual stream of LLMs. 2024. URLhttps://arxiv.org/abs/2409.17113. NeurIPS 2024 SciForDL Workshop

2024 arXiv
[72]

A structural probe for finding syntax in word representations

John Hewitt and Christopher D Manning. A structural probe for finding syntax in word representations. InProceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Pape...

2019
[73]

Fast weight programming and linear transformers: from machine learning to neurobiology.arXiv preprint arXiv:2508.08435, 2025

Kazuki Irie and Samuel J Gershman. Fast weight programming and linear transformers: from machine learning to neurobiology.arXiv preprint arXiv:2508.08435, 2025. URLhttps: //arxiv.org/abs/2508.08435

2025
[75]

URLhttps://arxiv.org/abs/2512.19941
[76]

Language models (mostly) know what they know.arXiv preprint arXiv:2207.05221, 2022

Saurav Kadavath, Tom Conerly, Amanda Askell, Tom Henighan, Dawn Drain, Ethan Perez, Nicholas Schiefer, Zac Hatfield-Dodds, Nova DasSarma, Eli Tran-Johnson, et al. Language models (mostly) know what they know.arXiv preprint arXiv:2207.05221, 2022. URL https://arxiv.org/pdf/2207.05221

2022 arXiv
[77]

macmillan, 2011

Daniel Kahneman.Thinking, Fast and Slow. macmillan, 2011

2011
[78]

Lan- guage models are capable of metacognitive monitoring and control of their internal activa- tions.arXiv preprint arXiv:2505.13763, 2025

Li Ji-An, Hua-Dong Xiong, Robert C Wilson, Marcelo G Mattar, and Marcus K Benna. Lan- guage models are capable of metacognitive monitoring and control of their internal activa- tions.arXiv preprint arXiv:2505.13763, 2025

2025
[80]

Backward lens: Projecting lan- guage model gradients into the vocabulary space

Shahar Katz, Yonatan Belinkov, Mor Geva, and Lior Wolf. Backward lens: Projecting lan- guage model gradients into the vocabulary space. InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, 2024. URLhttps://arxiv.org/ abs/2402.12865

2024 arXiv
[81]

Devils in middle layers of large vision-language models: Interpreting, detecting and mitigating object hallucinations via attention lens

Zhangqi Jiang, Junkai Chen, Beier Zhu, Tingjin Luo, Yankun Shen, and Xu Yang. Devils in middle layers of large vision-language models: Interpreting, detecting and mitigating object hallucinations via attention lens. InProceedings of the IEEE/CVF Conference on Computer Vision a...

2025 arXiv
[82]

Triviaqa: A large scale distantly supervised challenge dataset for reading comprehension.arXiv preprint arXiv:1705.03551, 2017

Mandar Joshi, Eunsol Choi, Daniel S Weld, and Luke Zettlemoyer. Triviaqa: A large scale distantly supervised challenge dataset for reading comprehension.arXiv preprint arXiv:1705.03551, 2017. URLhttps://arxiv.org/pdf/1705.03551

2017 arXiv
[83]

Building transformers from neu- rons and astrocytes.Proceedings of the National Academy of Sciences, 120(34):e2219150120,

Leo Kozachkov, Ksenia V Kastanenka, and Dmitry Krotov. Building transformers from neu- rons and astrocytes.Proceedings of the National Academy of Sciences, 120(34):e2219150120,
[84]

Atp*: An efficient and scalable method for localizing llm behaviour to components.arXiv preprint arXiv:2403.00745, 2024

János Kramár, Tom Lieberum, Rohin Shah, and Neel Nanda. Atp*: An efficient and scalable method for localizing llm behaviour to components.arXiv preprint arXiv:2403.00745, 2024. URLhttps://arxiv.org/pdf/2403.00745

2024 arXiv
[85]

Tracing attention computation through feature interactions.Transformer Circuits Thread,

Harish Kamath, Emmanuel Ameisen, Isaac Kauvar, Rodrigo Luger, Wes Gurnee, Adam Pearce, Sam Zimmerman, Joshua Batson, Thomas Conerly, Chris Olah, and Jack Lindsey. Tracing attention computation through feature interactions.Transformer Circuits Thread,
[86]

URLhttps://transformer-circuits.pub/2025/attention-qk/index.html

2025
[87]

Activation oracles: Training and evaluating llms as general-purpose activation explainers

Adam Karvonen, James Chua, Clément Dumas, Kit Fraser-Taliente, Subhash Kantamneni, Julian Minder, Euan Ong, Arnab Sen Sharma, Daniel Wen, Owain Evans, and Samuel Marks. Activation oracles: Training and evaluating llms as general-purpose activation explainers. arXiv preprint ar...

2025
[88]

Me, myself, and ai: The situational awareness dataset (sad) for llms.Advances in Neural Information Processing Systems, 37:64010–64118, 2024

Rudolf Laine, Bilal Chughtai, Jan Betley, Kaivalya Hariharan, Mikita Balesni, Jérémy Scheurer, Marius Hobbhahn, Alexander Meinke, and Owain Evans. Me, myself, and ai: The situational awareness dataset (sad) for llms.Advances in Neural Information Processing Systems, 37:64010–6...

2024
[89]

The cognitive unconscious.Science, 237(4821):1445–1452, 1987

John F Kihlstrom. The cognitive unconscious.Science, 237(4821):1445–1452, 1987

1987
[90]

Similarity of neural network representations revisited

Simon Kornblith, Mohammad Norouzi, Honglak Lee, and Geoffrey Hinton. Similarity of neural network representations revisited. InInternational conference on machine learning, pages 3519–3529. PMlR, 2019

2019
[91]

Empirical support for higher-order theories of conscious awareness.Trends in cognitive sciences, 15(8):365–373, 2011

Hakwan Lau and David Rosenthal. Empirical support for higher-order theories of conscious awareness.Trends in cognitive sciences, 15(8):365–373, 2011

2011
[92]

doi: 10.1073/pnas.2219150120

doi
[93]

Self unbound: ego dissolution in psychedelic experience

Chris Letheby and Philip Gerrans. Self unbound: ego dissolution in psychedelic experience. Neuroscience of Consciousness, 2017(1):nix016, 2017. doi: 10.1093/nc/nix016

2017 doi
[94]

LatentLens: Revealing highly interpretable visual tokens in LLMs

Benno Krojer, Shravan Nayak, Oscar Mañas, Vaibhav Adlakha, Desmond Elliott, Siva Reddy, and Marius Mosbach. LatentLens: Revealing highly interpretable visual tokens in LLMs. arXiv preprint arXiv:2602.00462, 2026. URLhttps://arxiv.org/abs/2602.00462

2026 arXiv
[95]

Large associative memory problem in neurobiology and machine learning

Dmitry Krotov and John J Hopfield. Large associative memory problem in neurobiology and machine learning. InInternational Conference on Learning Representations, 2021. URL https://arxiv.org/abs/2008.06996

2021 arXiv
[96]

The remarkable robustness of llms: Stages of inference?arXiv preprint arXiv:2406.19384, 2024

Vedang Lad, Jin Hwa Lee, Wes Gurnee, and Max Tegmark. The remarkable robustness of llms: Stages of inference?arXiv preprint arXiv:2406.19384, 2024

2024 arXiv
[97]

Emergent introspective awareness in large language models.Trans- former Circuits Thread, 2025

Jack Lindsey. Emergent introspective awareness in large language models.Trans- former Circuits Thread, 2025. URLhttps://transformer-circuits.pub/2025/ introspection/index.html

2025
[98]

Victor A. F. Lamme. Towards a true neural stance on consciousness.Trends in Cognitive Sciences, 10(11):494–501, 2006. doi: 10.1016/j.tics.2006.09.001

2006 doi
[99]

Victor A. F. Lamme and Pieter R. Roelfsema. The distinct modes of vision offered by feed- forward and recurrent processing.Trends in Neurosciences, 23(11):571–579, 2000. doi: 10.1016/S0166-2236(00)01657-X. 111

doi
[100]

Introspective capabilities in large language models.Journal of Consciousness Studies, 30(9-10):143–153, 2023

Robert Long. Introspective capabilities in large language models.Journal of Consciousness Studies, 30(9-10):143–153, 2023

2023
[101]

Sparse autoencoders do not find canonical units of analysis.arXiv preprint arXiv:2502.04878, 2025

Patrick Leask, Bart Bussmann, Michael Pearce, Joseph Bloom, Curt Tigges, Noura Al Moubayed, Lee Sharkey, and Neel Nanda. Sparse autoencoders do not find canonical units of analysis.arXiv preprint arXiv:2502.04878, 2025. URLhttps://arxiv.org/pdf/2502. 04878

2025 arXiv
[102]

Paths not taken: Under- standing and mending the multilingual factual recall pipeline

Meng Lu, Ruochen Zhang, Carsten Eickhoff, and Ellie Pavlick. Paths not taken: Under- standing and mending the multilingual factual recall pipeline. InProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 15066–15096,

2025
[103]

Inference-time intervention: Eliciting truthful answers from a language model, 2023

Kenneth Li, Oam Patel, Fernanda Viégas, Hanspeter Pfister, and Martin Wattenberg. Inference-time intervention: Eliciting truthful answers from a language model, 2023. URL https://arxiv.org/pdf/2306.03341

2023 arXiv
[104]

Understanding and patching compositional reasoning in llms

Zhaoyi Li, Gangwei Jiang, Hong Xie, Linqi Song, Defu Lian, and Ying Wei. Understanding and patching compositional reasoning in llms. InFindings of the Association for Computa- tional Linguistics: ACL 2024, pages 9668–9688, 2024. URLhttps://arxiv.org/abs/ 2402.14328

2024 arXiv
[105]

The hidden life of tokens: Reducing hal- lucination of large vision-language models via visual information steering.arXiv preprint arXiv:2502.03628, 2025

Zhuowei Li, Haizhou Shi, Yunhe Gao, Di Liu, Zhenting Wang, Yuxiao Chen, Ting Liu, Long Zhao, Hao Wang, and Dimitris N Metaxas. The hidden life of tokens: Reducing hal- lucination of large vision-language models via visual information steering.arXiv preprint arXiv:2502.03628, 2...

2025 arXiv
[106]

Auditing language models for hidden objectives

Samuel Marks, Johannes Treutlein, Trenton Bricken, Jack Lindsey, Jonathan Marcus, Sid- dharth Mishra-Sharma, Daniel Ziegler, Emmanuel Ameisen, Joshua Batson, Shan Carter, Brian Chen, Hoagy Cunningham, Florien Dietz, Satvik Golechha, Jan Kirchner, Austin Meek, Kei Nishimura-Gas...

2025 arXiv
[107]

Jack Lindsey, Wes Gurnee, Emmanuel Ameisen, Brian Chen, Adam Pearce, Nicholas L. Turner, Craig Citro, David Abrahams, Shan Carter, Basil Hosmer, Jonathan Marcus, Michael Sklar, Adly Templeton, Trenton Bricken, Callum McDougall, Hoagy Cunningham, Thomas Henighan, Adam Jermyn, A...

2025
[108]

Lin- guistic knowledge and transferability of contextual representations

Nelson F Liu, Matt Gardner, Yonatan Belinkov, Matthew E Peters, and Noah A Smith. Lin- guistic knowledge and transferability of contextual representations. InProceedings of the 2019 Conference of the North American Chapter of the Association for Computational Lin- guistics: Hu...

2019
[109]

Language models implement simple word2vec-style vector arithmetic

Jack Merullo, Carsten Eickhoff, and Ellie Pavlick. Language models implement simple word2vec-style vector arithmetic. InProceedings of the 2024 Conference of the North Amer- ican Chapter of the Association for Computational Linguistics: Human Language Technolo- gies, pages 503...

2024 arXiv
[110]

The assis- tant axis: Situating and stabilizing the default persona of language models.arXiv preprint arXiv:2601.10387, 2026

Christina Lu, Jack Gallagher, Jonathan Michala, Kyle Fish, and Jack Lindsey. The assis- tant axis: Situating and stabilizing the default persona of language models.arXiv preprint arXiv:2601.10387, 2026

2026
[111]

The magical number seven, plus or minus two: Some limits on our capacity for processing information.Psychological Review, 63(2):81–97, 1956

George A Miller. The magical number seven, plus or minus two: Some limits on our capacity for processing information.Psychological Review, 63(2):81–97, 1956. doi: 10.1037/h0043158

1956 doi
[112]

URLhttps://arxiv.org/abs/2505.20546

arXiv
[113]

Agentic misalignment: How llms could be insider threats.arXiv preprint arXiv:2510.05179, 2025

Aengus Lynch, Benjamin Wright, Caleb Larson, Stuart J Ritchie, Soren Mindermann, Evan Hubinger, Ethan Perez, and Kevin Troy. Agentic misalignment: How llms could be insider threats.arXiv preprint arXiv:2510.05179, 2025

2025
[114]

Natural emergent misalignment from reward hacking in production rl.arXiv preprint arXiv:2511.18397, 2025

Monte MacDiarmid, Benjamin Wright, Jonathan Uesato, Joe Benton, Jon Kutasov, Sara Price, Naia Bouscal, Sam Bowman, Trenton Bricken, Alex Cloud, et al. Natural emergent misalignment from reward hacking in production rl.arXiv preprint arXiv:2511.18397, 2025. 112

2025
[115]

Sparse feature circuits: Discovering and editing interpretable causal graphs in lan- guage models.arXiv preprint arXiv:2403.19647, 2024

Samuel Marks, Can Rager, Eric J Michaud, Yonatan Belinkov, David Bau, and Aaron Mueller. Sparse feature circuits: Discovering and editing interpretable causal graphs in lan- guage models.arXiv preprint arXiv:2403.19647, 2024. URLhttps://arxiv.org/pdf/ 2403.19647

2024 arXiv
[116]

Attribution patching: Activation patching at industrial scale, 2023

Neel Nanda. Attribution patching: Activation patching at industrial scale, 2023. URLhttps: //www.neelnanda.io/mechanistic-interpretability/attribution-patching

2023
[117]

Capacity limits of information processing in the brain.Trends in cognitive sciences, 9(6):296–305, 2005

René Marois and Jason Ivanoff. Capacity limits of information processing in the brain.Trends in cognitive sciences, 9(6):296–305, 2005

2005
[118]

Con- scious processing and the global neuronal workspace hypothesis.Neuron, 105(5):776–798,

George A Mashour, Pieter Roelfsema, Jean-Pierre Changeux, and Stanislas Dehaene. Con- scious processing and the global neuronal workspace hypothesis.Neuron, 105(5):776–798,
[119]

Show your work: Scratchpads for intermediate computation with language models.arXiv preprint arXiv:2112.00114, 2021

Maxwell Nye, Anders Johan Andreassen, Guy Gur-Ari, Henryk Michalewski, Jacob Austin, David Bieber, David Dohan, Aitor Lewkowycz, Maarten Bosma, David Luan, Charles Sut- ton, and Augustus Odena. Show your work: Scratchpads for intermediate computation with language models.arXiv...

2021 arXiv
[120]

In-context learning and induction heads

Catherine Olsson, Nelson Elhage, Neel Nanda, Nicholas Joseph, Nova DasSarma, Tom Henighan, Ben Mann, Amanda Askell, Yuntao Bai, Anna Chen, Tom Conerly, Dawn Drain, Deep Ganguli, Zac Hatfield-Dodds, Danny Hernandez, Scott Johnston, Andy Jones, Jackson Kernion, Liane Lovitt, Kam...

2022
[121]

Minimal phenomenal experience: Meditation, tonic alertness, and the phenomenology of "pure" consciousness.Philosophy and the Mind Sciences, 1(I):1–44, 2020

Thomas Metzinger. Minimal phenomenal experience: Meditation, tonic alertness, and the phenomenology of "pure" consciousness.Philosophy and the Mind Sciences, 1(I):1–44, 2020. doi: 10.33735/phimisci.2020.I.46

2020 doi
[123]

The varieties of selflessness.Philosophy and the Mind Sciences, 1(I):1–41,

Raphaël Millière. The varieties of selflessness.Philosophy and the Mind Sciences, 1(I):1–41,
[124]

doi: 10.33735/phimisci.2020.I.48

2020 doi
[125]

Synaptic theory of working memory

Gianluigi Mongillo, Omri Barak, and Misha Tsodyks. Synaptic theory of working memory. Science, 319(5869):1543–1546, 2008. doi: 10.1126/science.1150769

2008 doi
[126]

On the number of linear regions of deep neural networks

Guido Montúfar, Razvan Pascanu, Kyunghyun Cho, and Yoshua Bengio. On the number of linear regions of deep neural networks. InAdvances in Neural Information Processing Systems, 2014. URLhttps://arxiv.org/abs/1402.1869

2014 arXiv
[127]

Information integration without awareness

Liad Mudrik, Nathan Faivre, and Christof Koch. Information integration without awareness. Trends in cognitive sciences, 18(9):488–496, 2014

2014
[128]

Blindsight is unlike normal conscious vision: evidence from an exclusion task.Consciousness and cognition, 17(3):1050–1055, 2008

Navindra Persaud and Alan Cowey. Blindsight is unlike normal conscious vision: evidence from an exclusion task.Consciousness and cognition, 17(3):1050–1055, 2008

2008
[129]

Progress update #1 from the gdm mech interp team, 2024

Neel Nanda, Arthur Conmy, Lewis Smith, Senthooran Rajamanoharan, Tom Lieberum, János Kramár, and Vikrant Varma. Progress update #1 from the gdm mech interp team, 2024. URLhttps://www.alignmentforum.org/posts/HpAr8k74mW4ivCvCu/ summary-progress-update-1-from-the-gdm-mech-interp-team

2024
[130]

interpreting gpt: the logit lens

nostalgebraist. interpreting gpt: the logit lens. LessWrong, 2020. URLhttps://www. lesswrong.com/posts/AcKRB8wDpdaN6v6ru/interpreting-gpt-the-logit-lens

2020
[131]

Clarendon Press, 2005

David Rosenthal.Consciousness and mind. Clarendon Press, 2005

2005
[132]

The cognitive architecture for chaining of two mental operations.Cognition, 111(2):187–211, 2009

Jérôme Sackur and Stanislas Dehaene. The cognitive architecture for chaining of two mental operations.Cognition, 111(2):187–211, 2009

2009
[133]

A sensorimotor account of vision and visual consciousness

J Kevin O’regan and Alva Noë. A sensorimotor account of vision and visual consciousness. Behavioral and brain sciences, 24(5):939–973, 2001

2001
[134]

Linear transformers are secretly fast weight programmers

Imanol Schlag, Kazuki Irie, and Jürgen Schmidhuber. Linear transformers are secretly fast weight programmers. InProceedings of the 38th International Conference on Machine Learn- ing. PMLR, 2021. URLhttps://arxiv.org/abs/2102.11174

2021 arXiv
[135]

URLhttps://arxiv.org/pdf/2311.04897

arXiv
[136]

Latentqa: Teaching llms to decode ac- tivations into natural language.arXiv preprint arXiv:2412.08686, 2024

Alexander Pan, Lijie Chen, and Jacob Steinhardt. Latentqa: Teaching llms to decode ac- tivations into natural language.arXiv preprint arXiv:2412.08686, 2024. URLhttps: //arxiv.org/abs/2412.08686

2024
[137]

Steering llama 2 via contrastive activation addition, 2024.URL https://arxiv

Nina Panickssery, Nick Gabrieli, Julian Schulz, Meg Tong, Evan Hubinger, and Alexan- der Matt Turner. Steering llama 2 via contrastive activation addition, 2024.URL https://arxiv. org/abs/2312.06681, 3

2024 arXiv
[138]

The linear representation hypothesis and the geometry of large language models.arXiv preprint arXiv:2311.03658, 2023

Kiho Park, Yo Joong Choe, and Victor Veitch. The linear representation hypothesis and the geometry of large language models.arXiv preprint arXiv:2311.03658, 2023

2023 arXiv
[139]

Automatically interpreting millions of features in large language models.arXiv preprint arXiv:2410.13928, 2024

Gonçalo Paulo, Alex Mallen, Caden Juang, and Nora Belrose. Automatically interpreting millions of features in large language models.arXiv preprint arXiv:2410.13928, 2024. URL https://arxiv.org/pdf/2410.13928

2024 arXiv
[140]

Transcoders beat sparse autoencoders for interpretability.arXiv preprint arXiv:2501.18823, 2025

Gonçalo Paulo, Stepan Shabalin, and Nora Belrose. Transcoders beat sparse autoencoders for interpretability.arXiv preprint arXiv:2501.18823, 2025. URLhttps://arxiv.org/pdf/ 2501.18823

2025 arXiv
[141]

Privileged self-access matters for introspection in ai.arXiv preprint arXiv:2508.14802, 2025

Siyuan Song, Harvey Lederman, Jennifer Hu, and Kyle Mahowald. Privileged self-access matters for introspection in ai.arXiv preprint arXiv:2508.14802, 2025

2025 arXiv
[142]

Self-interpretability: Llms can describe complex internal processes that drive their decisions, and improve with training

Dillon Plunkett, Adam Morris, Keerthi Reddy, and Jorge Morales. Self-interpretability: Llms can describe complex internal processes that drive their decisions, and improve with training. arXiv preprint arXiv:2505.17120, 2025

2025
[143]

Squad: 100,000+ ques- tions for machine comprehension of text

Pranav Rajpurkar, Jian Zhang, Konstantin Lopyrev, and Percy Liang. Squad: 100,000+ ques- tions for machine comprehension of text. InProceedings of the 2016 Conference on Em- pirical Methods in Natural Language Processing (EMNLP), pages 2383–2392, 2016. URL https://arxiv.org/ab...

2016 arXiv
[144]

Improving instruction-following in language models through activa- tion steering

Alessandro Stolfo et al. Improving instruction-following in language models through activa- tion steering. InarXiv preprint arXiv:2410.12877, 2024. URLhttps://arxiv.org/abs/ 2410.12877

2024 arXiv
[145]

Attribution patching outperforms automated circuit discovery.arXiv preprint arXiv:2310.10348, 2023

Aaquib Syed, Can Rager, and Arthur Conmy. Attribution patching outperforms automated circuit discovery.arXiv preprint arXiv:2310.10348, 2023. URLhttps://arxiv.org/pdf/ 2310.10348

2023 arXiv
[146]

Romanlens: The role of latent romanization in multilinguality in llms

Alan Saji, Jaavid Aktar Husain, Thanmay Jayakumar, Raj Dabre, Anoop Kunchukuttan, and Ratish Puduppully. Romanlens: The role of latent romanization in multilinguality in llms. In Findings of the Association for Computational Linguistics: ACL 2025, 2025. URLhttps: //arxiv.org/a...

2025
[147]

Daniel Freeman, Theodore R

Adly Templeton, Tom Conerly, Jonathan Marcus, Jack Lindsey, Trenton Bricken, Brian Chen, Adam Pearce, Craig Citro, Emmanuel Ameisen, Andy Jones, Hoagy Cunningham, Nicholas L Turner, Callum McDougall, Monte MacDiarmid, C. Daniel Freeman, Theodore R. Sumers, Edward Rees, Joshua ...

2024
[148]

Do multilingual llms think in english?, 2025

Lisa Schut, Yarin Gal, and Sebastian Farquhar. Do multilingual llms think in english?, 2025. URLhttps://arxiv.org/pdf/2502.15603

2025 arXiv
[149]

MIT press, 1992

John R Searle.The rediscovery of the mind. MIT press, 1992. 114

1992
[150]

Conscious artificial intelligence and biological naturalism.Behavioral and Brain Sciences, pages 1–42, 2024

Anil K Seth. Conscious artificial intelligence and biological naturalism.Behavioral and Brain Sciences, pages 1–42, 2024

2024
[151]

Deep inside convolutional networks: Visualising image classification models and saliency maps.arXiv preprint arXiv:1312.6034, 2013

Karen Simonyan, Andrea Vedaldi, and Andrew Zisserman. Deep inside convolutional networks: Visualising image classification models and saliency maps.arXiv preprint arXiv:1312.6034, 2013. URLhttps://arxiv.org/pdf/1312.6034

2013 arXiv
[152]

Layer by layer: Uncovering hidden representations in language models

Oscar Skean, Md Rifat Arefin, Dan Zhao, Niket Patel, Jalal Naghiyev, Yann LeCun, and Ravid Shwartz-Ziv. Layer by layer: Uncovering hidden representations in language models. InProceedings of the 42nd International Conference on Machine Learning, 2025. URL https://arxiv.org/abs...

2025 arXiv
[153]

Emotion concepts and their function in a large language model.Transformer Circuits Thread, 2026

Nicholas Sofroniew, Isaac Kauvar, William Saunders, Runjin Chen, Tom Henighan, Sasha Hydrie, Craig Citro, Adam Pearce, Julius Tarng, Wes Gurnee, Joshua Batson, Sam Zim- merman, Kelley Rivoire, Kyle Fish, Chris Olah, and Jack Lindsey. Emotion concepts and their function in a la...

2026
[154]

Rosas, Jun Wang, Andrea I

Pedro Urbina-Rodriguez, Zafeirios Fountas, Fernando E. Rosas, Jun Wang, Andrea I. Luppi, Haitham Bou-Ammar, Murray Shanahan, and Pedro A. M. Mediano. A brain-like synergistic core in LLMs drives behaviour and learning. 2026. URLhttps://arxiv.org/abs/2601. 06851

2026
[155]

Ziegler, Ryan Lowe, Chelsea V oss, Alec Radford, Dario Amodei, and Paul Christiano

Nisan Stiennon, Long Ouyang, Jeff Wu, Daniel M. Ziegler, Ryan Lowe, Chelsea V oss, Alec Radford, Dario Amodei, and Paul Christiano. Learning to summarize from human feedback. InAdvances in Neural Information Processing Systems (NeurIPS), 2020. URLhttps:// arxiv.org/abs/2009.01325

2020 arXiv
[156]

‘activity-silent’ working memory in prefrontal cortex: a dynamic coding framework.Trends in Cognitive Sciences, 19(7):394–405, 2015

Mark G Stokes. ‘activity-silent’ working memory in prefrontal cortex: a dynamic coding framework.Trends in Cognitive Sciences, 19(7):394–405, 2015. doi: 10.1016/j.tics.2015.05. 004

2015 doi
[157]

Tl;dr: Mining Reddit to learn automatic summarization

Michael Völske, Martin Potthast, Shahbaz Syed, and Benno Stein. Tl;dr: Mining Reddit to learn automatic summarization. InProceedings of the Workshop on New Frontiers in Summarization, pages 59–63. Association for Computational Linguistics, 2017. doi: 10. 18653/v1/W17-4508. URL...

2017
[158]

Interpretability in the wild: a circuit for indirect object identification in gpt-2 small.arXiv preprint arXiv:2211.00593, 2022

Kevin Wang, Alexandre Variengien, Arthur Conmy, Buck Shlegeris, and Jacob Steinhardt. Interpretability in the wild: a circuit for indirect object identification in gpt-2 small.arXiv preprint arXiv:2211.00593, 2022. URLhttps://arxiv.org/pdf/2211.00593

2022 arXiv
[159]

Decoding vision trans- formers: the diffusion steering lens

Ryota Takatsuki, Sonia Joseph, Ippei Fujisawa, and Ryota Kanai. Decoding vision trans- formers: the diffusion steering lens. InCVPR Workshop on Mechanistic Interpretability for Vision, 2025. URLhttps://arxiv.org/abs/2504.13763

2025 arXiv
[160]

Ironic processes of mental control.Psychological Review, 101(1):34–52,

Daniel M Wegner. Ironic processes of mental control.Psychological Review, 101(1):34–52,
[161]

Bert rediscovers the classical nlp pipeline

Ian Tenney, Dipanjan Das, and Ellie Pavlick. Bert rediscovers the classical nlp pipeline. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4593–4601, Florence, Italy, 2019. Association for Computational Linguistics. doi: 10.1865...

2019 doi
[162]

Linear representa- tions of sentiment in large language models, 2023

Curt Tigges, Oskar John Hollinsworth, Atticus Geiger, and Neel Nanda. Linear representa- tions of sentiment in large language models, 2023. URLhttps://arxiv.org/pdf/2310. 15154

2023
[163]

Function vectors in large language models.arXiv preprint arXiv:2310.15213, 2023

Eric Todd, Millicent L Li, Arnab Sen Sharma, Aaron Mueller, Byron C Wallace, and David Bau. Function vectors in large language models.arXiv preprint arXiv:2310.15213, 2023. URLhttps://arxiv.org/pdf/2310.15213. 115

2023 arXiv
[164]

Diffusion lens: Interpreting text encoders in text-to-image pipelines

Michael Toker, Hadas Orgad, Mor Ventura, Dana Arad, and Yonatan Belinkov. Diffusion lens: Interpreting text encoders in text-to-image pipelines. InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 9713–9728, ...

2024 arXiv
[165]

What language(s) does aya-23 think in? how multilinguality affects internal language representations.arXiv preprint arXiv:2507.20279, 2025

Katharina Trinley, Toshiki Nakai, Tatiana Anikina, and Tanja Baeumel. What language(s) does aya-23 think in? how multilinguality affects internal language representations.arXiv preprint arXiv:2507.20279, 2025. URLhttps://arxiv.org/abs/2507.20279

2025 arXiv
[166]

Activation addition: Steering language models without optimization, 2023

Alexander Matt Turner, Lisa Thiergart, David Udell, Gavin Leech, Ulisse Mini, and Monte MacDiarmid. Activation addition: Steering language models without optimization, 2023. URLhttps://arxiv.org/pdf/2308.10248

2023 arXiv
[167]

Jump to conclusions: Short-cutting transformers with linear transformations

Alexander Yom Din, Taelin Karidi, Leshem Choshen, and Mor Geva. Jump to conclusions: Short-cutting transformers with linear transformations. InProceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 20...

2024
[168]

The geometry of hidden representations of large transformer models

Lucrezia Valeriani, Diego Doimo, Francesca Cuturello, Alessandro Laio, Alessio Ansuini, and Alberto Cazzaniga. The geometry of hidden representations of large transformer models. InAdvances in Neural Information Processing Systems, 2023. URLhttps://arxiv.org/ abs/2302.00294

2023 arXiv
[169]

Deep learning and the global workspace theory.Trends in Neurosciences, 44(9):692–704, 2021

Rufin VanRullen and Ryota Kanai. Deep learning and the global workspace theory.Trends in Neurosciences, 44(9):692–704, 2021. doi: 10.1016/j.tins.2021.04.005. URLhttps:// arxiv.org/abs/2012.10390

2021 arXiv
[170]

How alignment and jailbreak work: Explain llm safety through intermediate hidden states

Zhenhong Zhou, Haiyang Yu, Xinghua Zhang, Rongwu Xu, Fei Huang, and Yongbin Li. How alignment and jailbreak work: Explain llm safety through intermediate hidden states. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 2461–2488,

2024
[171]

Representation engi- neering: A top-down approach to ai transparency.arXiv preprint arXiv:2310.01405, 2023

Andy Zou, Long Phan, Sarah Chen, James Campbell, Phillip Guo, Richard Ren, Alexander Pan, Xuwang Yin, Mantas Mazeika, Ann-Kathrin Dombrowski, et al. Representation engi- neering: A top-down approach to ai transparency.arXiv preprint arXiv:2310.01405, 2023. URLhttps://arxiv.org...

2023 arXiv
[172]

Alex Warstadt, Amanpreet Singh, and Samuel R. Bowman. Neural network acceptability judgments.Transactions of the Association for Computational Linguistics, 7:625–641, 2019. URLhttps://arxiv.org/abs/1805.12471

2019 arXiv
[175]

Paradoxical effects of thought suppression.Journal of Personality and Social Psychology, 53(1):5–13, 1987

Daniel M Wegner, David J Schneider, Samuel R Carter, and Teri L White. Paradoxical effects of thought suppression.Journal of Personality and Social Psychology, 53(1):5–13, 1987. doi: 10.1037/0022-3514.53.1.5

1987 doi
[176]

Chi, Quoc V

Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed H. Chi, Quoc V . Le, and Denny Zhou. Chain-of-thought prompting elicits reasoning in large language models. InAdvances in Neural Information Processing Systems (NeurIPS), 2022. URLhttps://arxiv.o...

2022 arXiv
[177]

Oxford University Press, 1986

Lawrence Weiskrantz.Blindsight: A Case Study and Implications. Oxford University Press, 1986

1986
[178]

Do llamas work in english? on the latent language of multilingual transformers

Chris Wendler, Veniamin Veselovsky, Giovanni Monea, and Robert West. Do llamas work in english? on the latent language of multilingual transformers. InProceedings of the 62nd An- nual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 15366...

2024
[179]

Addition in four movements: Mapping layer-wise information trajectories in llms

Yao Yan. Addition in four movements: Mapping layer-wise information trajectories in llms. InFindings of the Association for Computational Linguistics: EMNLP 2025, 2025. URL https://arxiv.org/abs/2506.07824. 116

2025 arXiv
[181]

URLhttps://arxiv.org/pdf/2402.16837

arXiv
[183]

Knowledge overshadowing causes amalgamated hallucination in large language models

Yuji Zhang, Sha Li, Jiateng Liu, Pengfei Yu, Yi R Fung, Jing Li, Manling Li, and Heng Ji. Knowledge overshadowing causes amalgamated hallucination in large language models. arXiv preprint arXiv:2407.08039, 2024. URLhttps://arxiv.org/abs/2407.08039

2024 arXiv
[184]

Beyond english-centric llms: What language do multi- lingual language models think in?arXiv preprint arXiv:2408.10811, 2024

Chengzhi Zhong, Fei Cheng, Qianying Liu, Junfeng Jiang, Zhen Wan, Chenhui Chu, Yugo Murawaki, and Sadao Kurohashi. Beyond english-centric llms: What language do multi- lingual language models think in?arXiv preprint arXiv:2408.10811, 2024. URLhttps: //arxiv.org/abs/2408.10811

2024 arXiv
[186]

URLhttps://arxiv.org/abs/2406.05644

arXiv
[1994]

doi: 10.1037/0033-295X.101.1.34

doi
[2020]

doi: 10.1016/j.neuron.2020.01.026

2020 doi
[2023]

URLhttps://arxiv.org/pdf/2309.08600

arXiv
[2024]

URLhttps://arxiv.org/abs/2410.02762

arXiv
[2025]

URLhttps://proceedings.neurips.cc/paper_files/paper/2024/file/ f545448535dfde4f9786555403ab7c49-Paper-Conference.pdf

2024

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [1]

Understanding intermediate layers using linear clas- sifier probes.arXiv preprint arXiv:1610.01644, 2016

Guillaume Alain and Yoshua Bengio. Understanding intermediate layers using linear clas- sifier probes.arXiv preprint arXiv:1610.01644, 2016. URLhttps://arxiv.org/pdf/ 1610.01644

arXiv 2016

[2] [2]

Integrated information theory (iit) 4.0: Formulating the properties of phenomenal exis- tence in physical terms.PLoS computational biology, 19(10):e1011465, 2023

Larissa Albantakis, Leonardo Barbosa, Graham Findlay, Matteo Grasso, Andrew M Haun, William Marshall, William GP Mayner, Alireza Zaeemzadeh, Melanie Boly, Bjørn E Juel, et al. Integrated information theory (iit) 4.0: Formulating the properties of phenomenal exis- tence in physical terms.PLoS computational biology, 19(10):e1011465, 2023

2023

[3] [3]

Emmanuel Ameisen, Jack Lindsey, Adam Pearce, Wes Gurnee, Nicholas L. Turner, Brian Chen, Craig Citro, David Abrahams, Shan Carter, Basil Hosmer, Jonathan Marcus, Michael Sklar, Adly Templeton, Trenton Bricken, Callum McDougall, Hoagy Cunningham, Thomas Henighan, Adam Jermyn, Andy Jones, Andrew Persic, Zhenyi Qi, T. Ben Thompson, Sam Zimmerman, Kelley Rivo...

2025

[4] [4]

Claude opus 4.5 system card

Anthropic. Claude opus 4.5 system card. Technical report, Anthropic, November 2025. URL https://www.anthropic.com/claude-opus-4-5-system-card

2025

[5] [5]

Claude sonnet 4.5 system card

Anthropic. Claude sonnet 4.5 system card. Technical report, Anthropic, 2025. URLhttps: //www.anthropic.com/claude-sonnet-4-5-system-card

2025

[6] [6]

Refusal in language models is mediated by a single direction.Advances in Neural Information Processing Systems, 37:136037–136083,

Andy Arditi, Oscar Obeso, Aaquib Syed, Daniel Paleka, Nina Panickssery, Wes Gurnee, and Neel Nanda. Refusal in language models is mediated by a single direction.Advances in Neural Information Processing Systems, 37:136037–136083,

[7] [7]

Language model circuits are sparse in the neuron basis.arXiv preprint arXiv:2601.22594, 2026

Aryaman Arora, Zhengxuan Wu, Jacob Steinhardt, and Sarah Schwettmann. Language model circuits are sparse in the neuron basis.arXiv preprint arXiv:2601.22594, 2026

arXiv 2026

[8] [8]

Tensorlens: End-to-end transformer analysis via high-order attention tensors.arXiv preprint arXiv:2601.17958, 2026

Ido Andrew Atad, Itamar Zimerman, Shahar Katz, and Lior Wolf. Tensorlens: End-to-end transformer analysis via high-order attention tensors.arXiv preprint arXiv:2601.17958, 2026. URLhttps://arxiv.org/abs/2601.17958

arXiv 2026