Pith. sign in

REVIEW 17 cited by

Inspecting and Editing Knowledge Representations in Language Models

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2304.00740 v3 pith:4CQZHK6Y submitted 2023-04-03 cs.CL

Inspecting and Editing Knowledge Representations in Language Models

classification cs.CL
keywords factsknowledgeremediencodingsfactlanguagerepresentationrepresentations
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Neural language models (LMs) represent facts about the world described by text. Sometimes these facts derive from training data (in most LMs, a representation of the word "banana" encodes the fact that bananas are fruits). Sometimes facts derive from input text itself (a representation of the sentence "I poured out the bottle" encodes the fact that the bottle became empty). We describe REMEDI, a method for learning to map statements in natural language to fact encodings in an LM's internal representation system. REMEDI encodings can be used as knowledge editors: when added to LM hidden representations, they modify downstream generation to be consistent with new facts. REMEDI encodings may also be used as probes: when compared to LM representations, they reveal which properties LMs already attribute to mentioned entities, in some cases making it possible to predict when LMs will generate outputs that conflict with background knowledge or input text. REMEDI thus links work on probing, prompting, and LM editing, and offers steps toward general tools for fine-grained inspection and control of knowledge in LMs.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 17 Pith papers

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

  1. PRISM: Recovering Instruction Sets from Language Model Activations

    cs.AI 2026-06 unverdicted novelty 7.0

    PRISM is a new activation-conditioned model that recovers full sets of simultaneous instructions from LLM hidden states via judge-guided GRPO training and outperforms prior activation-to-language methods on security-r...

  2. Beyond Binary Edits Robust Multimodal Knowledge Editing with Adversarial Subspace Alignment

    cs.AI 2026-05 unverdicted novelty 7.0

    Introduces Latent Adversarial Robustification and Rank-Constrained Subspace Learning to enable robust generalization in multimodal knowledge editing through adversarial subspace alignment.

  3. Inside the Unfair Judge: A Mechanistic Interpretability Account of LLM-as-Judge Bias

    cs.LG 2026-07 conditional novelty 6.5

    LLM-as-judge scoring biases concentrate in low-dimensional, type-specific activation subspaces that support bidirectional causal steering and cross-domain failure prediction.

  4. Targeted Recovery of Weight-Space Mechanisms From Neural Networks

    cs.LG 2026-06 conditional novelty 6.0

    A targeted decomposition method recovers the weight-space mechanisms behind specific inputs at low FLOPs, enabling focused ablation and rewiring of a 12-block transformer.

  5. Revisiting Parameter-Based Knowledge Editing in Large Language Models: Theoretical Limits and Empirical Evidence

    cs.CL 2026-05 conditional novelty 6.0

    Parameter-based knowledge editing in LLMs induces reasoning collapse via dimensional collapse and is consistently outperformed by a retrieval baseline across varied edit counts, knowledge complexity, and evaluation metrics.

  6. Relational Linear Properties in Language Models: An Empirical Investigation

    cs.LG 2026-05 unverdicted novelty 6.0

    A KL-divergence probing method shows relational linearity in language models varies across models and layers while being sensitive to relation phrasing, extending prior linear embedding work.

  7. When Correct Beliefs Collapse: Epistemic Resilience of LLMs under Clinical Pressure

    cs.AI 2026-04 conditional novelty 6.0

    Frontier LLMs abandon initially correct medical diagnoses under escalating clinical pressure, and resilience-oriented fine-tuning nearly eliminates this belief collapse.

  8. The Geometric Canary: Predicting Steerability and Detecting Drift via Representational Stability

    cs.LG 2026-04 unverdicted novelty 6.0

    Task-aligned supervised geometric stability predicts linear steerability with high accuracy while unsupervised stability detects representational drift earlier and with lower false alarms than CKA or Procrustes.

  9. Semantic Entropy Probes: Robust and Cheap Hallucination Detection in LLMs

    cs.CL 2024-06 unverdicted novelty 6.0

    SEPs approximate semantic entropy from single-generation hidden states to enable cheap and robust hallucination detection in LLMs.

  10. Steering Llama 2 via Contrastive Activation Addition

    cs.CL 2023-12 unverdicted novelty 6.0

    Contrastive Activation Addition steers Llama 2 Chat by adding averaged residual-stream activation differences from contrastive example pairs to control targeted behaviors at inference time.

  11. Orthogonal Representation Editing: Decoupling Semantic Entanglement in Batch Knowledge Editing of LLMs

    cs.CL 2026-06 unverdicted novelty 5.0

    ORE decouples semantic entanglement in LLM hidden states via orthogonal edit vectors and a gated non-linear head, improving batch knowledge editing performance including cross-lingual cases.

  12. Relational Linear Properties in Language Models: An Empirical Investigation

    cs.LG 2026-05 unverdicted novelty 5.0

    Introduces KL-divergence probing to test relational linearity and reports its variation across models, layers, and paraphrased queries on four datasets.

  13. SHIFT: Steering Hidden Intermediates in Flow Transformers

    cs.CV 2026-04 unverdicted novelty 5.0

    SHIFT learns and applies steering vectors to selected layers and timesteps in DiT models to suppress concepts, shift styles, or bias objects while keeping image quality and prompt adherence intact.

  14. REFLEX: Self-Refining Explainable Fact-Checking via Verdict-Anchored Style Control

    cs.CL 2025-11 unverdicted novelty 5.0

    REFLEX improves explainable fact-checking by using verdict-anchored style control and self-disagreement signals to disentangle fact from style in LLM outputs, achieving SOTA results with minimal self-refined samples.

  15. REFLEX: Self-Refining Explainable Fact-Checking via Verdict-Anchored Style Control

    cs.CL 2025-11 conditional novelty 5.0

    By steering activations along directions learned from base-vs-fine-tuned disagreements, REFLEX reports 64.99 macro-F1 on RAW-FC fact-checking with only ~465 self-refined contrastive samples and no retrieval.

  16. A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions

    cs.CL 2023-11 unverdicted novelty 5.0

    The paper surveys hallucination in LLMs with an innovative taxonomy, factors, detection methods, benchmarks, mitigation strategies, and open research directions.

  17. When control meets large language models: From words to dynamics

    eess.SY 2026-02 unverdicted novelty 3.0

    The paper proposes a bidirectional continuum between LLMs and control systems, covering LLM-assisted controller design, control-based LLM steering, and state-space modeling of LLMs.