Pith. sign in

REVIEW 8 cited by

Knowledge Circuits in Pretrained Transformers

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 2405.17969 v4 pith:7MGMJGQQ submitted 2024-05-28 cs.CL cs.AIcs.CVcs.IRcs.LG

classification cs.CLcs.AIcs.CVcs.IRcs.LG
keywords knowledgecircuitseditinglanguagemodelmodelsheadsmultilayer
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

The remarkable capabilities of modern large language models are rooted in their vast repositories of knowledge encoded within their parameters, enabling them to perceive the world and engage in reasoning. The inner workings of how these models store knowledge have long been a subject of intense interest and investigation among researchers. To date, most studies have concentrated on isolated components within these models, such as the Multilayer Perceptrons and attention head. In this paper, we delve into the computation graph of the language model to uncover the knowledge circuits that are instrumental in articulating specific knowledge. The experiments, conducted with GPT2 and TinyLLAMA, have allowed us to observe how certain information heads, relation heads, and Multilayer Perceptrons collaboratively encode knowledge within the model. Moreover, we evaluate the impact of current knowledge editing techniques on these knowledge circuits, providing deeper insights into the functioning and constraints of these editing methodologies. Finally, we utilize knowledge circuits to analyze and interpret language model behaviors such as hallucinations and in-context learning. We believe the knowledge circuits hold potential for advancing our understanding of Transformers and guiding the improved design of knowledge editing. Code and data are available in https://github.com/zjunlp/KnowledgeCircuits.

Discussion (0). Sign in 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

  1. Is Random Attention Sufficient for Sequence Modeling? Disentangling Trainable Components in the Transformer

    cs.LG 2025-06 conditional novelty 7.0 of 10

    Freezing query and key attention weights still lets transformers form induction heads and stay close to standard performance on language modeling, while random static attention (MixiT) fails on in-context tasks but su...

  2. HALT: Verification-Aware Stopping for Retrieval-Augmented Search Agents

    cs.AI 2026-08 conditional novelty 6.0 of 10

    A per-hop evidence coverage verifier can stop multi-hop retrieval agents early, cutting search loops by up to 45% while preserving standardized-extractor exact match.

  3. COMPKE: Complex Question Answering under Knowledge Editing

    cs.CL 2025-06 conditional novelty 6.0 of 10

    COMPKE is a new benchmark with 11,924 complex questions that tests knowledge editing through one-to-many relations and logical operations, where existing editing methods often fail.

  4. Do We Know What LLMs Don't Know? A Study of Consistency in Knowledge Probing

    cs.CL 2025-05 conditional novelty 6.0 of 10

    Knowledge probing methods for LLMs are found to be highly inconsistent under minor prompt perturbations and across different probes.

  5. Paths Not Taken: Understanding and Mending the Multilingual Factual Recall Pipeline

    cs.CL 2025-05 conditional novelty 6.0 of 10

    LLMs recall facts through an English-centric internal path and then translate the answer; injecting a translation vector and a recall vector raises accuracy by over 35 percentage points in the weakest language.

  6. Pierce the Mists, Greet the Sky: Decipher Knowledge Overshadowing via Knowledge Circuit Analysis

    cs.CL 2025-05 conditional novelty 6.0 of 10

    PhantomCircuit traces knowledge overshadowing to attention circuits during training and prunes circuit edges to recover the overshadowed answer.

  7. GloSS over Toxicity: Understanding and Mitigating Toxicity in LLMs via Global Toxic Subspace

    cs.CL 2025-05 conditional novelty 6.0 of 10

    Detoxifying LLMs by deleting a global, cross-layer 'toxic subspace' from feed-forward weights reduces toxic outputs more than layer-local subspace methods.

  8. Unveiling Instruction-Specific Neurons & Experts: An Analytical Framework for LLM's Instruction-Following Capabilities

    cs.CL 2025-05 conditional novelty 4.0 of 10

    Activation-frequency analysis identifies sparse units in LLMs that respond to instructions; same-category instructions share more of these units than different-category ones, and fine-tuning measurably changes the sets.

Pith tools