POISE trains a lightweight probe on the actor's internal states to predict expected rewards for RLVR, matching DAPO performance on math benchmarks with lower compute by avoiding extra rollouts or critic models.
hub
A mathematical framework for transformer circuits.Transformer Circuits Thread
13 Pith papers cite this work. Polarity classification is still indexing.
13
Pith papers citing it
hub tools
citation-role summary
background 3
citation-polarity summary
roles
background 3representative citing papers
Transformer circuits show free evolution during SFT, rendering static mechanistic localization inadequate for future parameter updates due to inherent temporal latency.
In a controlled synthetic setting, transformers implement in-distribution task inference via convex combinations of task vectors and out-of-distribution inference via nearly orthogonal extrapolative representations.
Averaging and temporally interpolating text latents in VLAs enables 83% success on novel task combinations in the libero-ood benchmark where SOTA models achieve under 15%.
On Llama-2-7B, path-rerouting magnitude in paired transcoder attribution graphs correlates with jailbreak success (r=0.461), while static node metrics and top-feature ablations do not.
MAGS learns low-dimensional subspaces from correct versus incorrect reasoning traces and applies targeted projection corrections to attention heads when they deviate from the correctness manifold during inference.
Probe trajectories across token positions in LRMs, combined with signal-processing features, improve prediction of future model outputs over static probes on safety and math tasks.
Sparse autoencoders on EEG transformers extract clinical features, identify three steering regimes, expose age-pathology entanglements and wrecking-ball failures, and map interventions to frequency spectra.
LLMs exhibit three geometric phases in next-token prediction—seeding multiplexing, hoisting overriding, and focal convergence—where predictive subspaces rise, stabilize, and converge across layers.
Mechanistic analysis of six preference optimization methods reveals distinct geometric shifts in model representations, with KTO/GRPO enhancing separability while DPO/ORPO degrade it.
Transformers can be built to act as nonlinear featurizers via attention, supporting in-context regression with proven generalization bounds on synthetic tasks.
With shortcut-decorrelated training, bidirectional masking, and a corrective objective, depth scaling makes direct transformer reasoning over Horn clauses match CoT accuracy within the training horizon, but not beyond it.
RET learns temporally consistent macrovariables from LLM activations via self-supervised learning to support interpretability, early behavioral prediction, and causal intervention.
citing papers explorer
-
Your Language Model is Its Own Critic: Reinforcement Learning with Value Estimation from Actor's Internal States
POISE trains a lightweight probe on the actor's internal states to predict expected rewards for RLVR, matching DAPO performance on math benchmarks with lower compute by avoiding extra rollouts or critic models.
-
Navigating by Old Maps: The Pitfalls of Static Mechanistic Localization in LLM Post-Training
Transformer circuits show free evolution during SFT, rendering static mechanistic localization inadequate for future parameter updates due to inherent temporal latency.
-
Task Vector Geometry Underlies Dual Modes of Task Inference in Transformers
In a controlled synthetic setting, transformers implement in-distribution task inference via convex combinations of task vectors and out-of-distribution inference via nearly orthogonal extrapolative representations.
-
VLAs are Confined yet Capable of Generalizing to Novel Instructions
Averaging and temporally interpolating text latents in VLAs enables 83% success on novel task combinations in the libero-ood benchmark where SOTA models achieve under 15%.
-
Mechanistic Interpretability of LLM Jailbreaks via Internal Attribution Graphs
On Llama-2-7B, path-rerouting magnitude in paired transcoder attribution graphs correlates with jailbreak success (r=0.461), while static node metrics and top-feature ablations do not.
-
Manifold-Guided Attention Steering
MAGS learns low-dimensional subspaces from correct versus incorrect reasoning traces and applies targeted projection corrections to attention heads when they deviate from the correctness manifold during inference.
-
Monitoring the Internal Monologue: Probe Trajectories Reveal Reasoning Dynamics
Probe trajectories across token positions in LRMs, combined with signal-processing features, improve prediction of future model outputs over static probes on safety and math tasks.
-
Mechanistic Interpretability of EEG Foundation Models via Sparse Autoencoders
Sparse autoencoders on EEG transformers extract clinical features, identify three steering regimes, expose age-pathology entanglements and wrecking-ball failures, and map interventions to frequency spectra.
-
A Geometric Perspective on Next-Token Prediction in Large Language Models: Three Emerging Phases
LLMs exhibit three geometric phases in next-token prediction—seeding multiplexing, hoisting overriding, and focal convergence—where predictive subspaces rise, stabilize, and converge across layers.
-
Mechanistic Analysis of Alignment Algorithms in Language Models
Mechanistic analysis of six preference optimization methods reveals distinct geometric shifts in model representations, with KTO/GRPO enhancing separability while DPO/ORPO degrade it.
-
Understanding In-Context Learning for Nonlinear Regression with Transformers: Attention as Featurizer
Transformers can be built to act as nonlinear featurizers via attention, supporting in-context regression with proven generalization bounds on synthetic tasks.
-
The Scaling Properties of Implicit Deductive Reasoning in Transformers
With shortcut-decorrelated training, bidirectional masking, and a corrective objective, depth scaling makes direct transformer reasoning over Horn clauses match CoT accuracy within the training horizon, but not beyond it.
-
Towards Effective Theory of LLMs: A Representation Learning Approach
RET learns temporally consistent macrovariables from LLM activations via self-supervised learning to support interpretability, early behavioral prediction, and causal intervention.