Pith. sign in

REVIEW 19 cited by

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 2304.14997 v4 pith:5EI4AKGL submitted 2023-04-28 cs.LG

Towards Automated Circuit Discovery for Mechanistic Interpretability

classification cs.LG
keywords behaviorcircuitinterpretabilityacdccomponentdatasetgpt-2mechanistic
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Through considerable effort and intuition, several recent works have reverse-engineered nontrivial behaviors of transformer models. This paper systematizes the mechanistic interpretability process they followed. First, researchers choose a metric and dataset that elicit the desired model behavior. Then, they apply activation patching to find which abstract neural network units are involved in the behavior. By varying the dataset, metric, and units under investigation, researchers can understand the functionality of each component. We automate one of the process' steps: to identify the circuit that implements the specified behavior in the model's computational graph. We propose several algorithms and reproduce previous interpretability results to validate them. For example, the ACDC algorithm rediscovered 5/5 of the component types in a circuit in GPT-2 Small that computes the Greater-Than operation. ACDC selected 68 of the 32,000 edges in GPT-2 Small, all of which were manually found by previous work. Our code is available at https://github.com/ArthurConmy/Automatic-Circuit-Discovery.

discussion (0)

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

Forward citations

Cited by 19 Pith papers

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

  1. Dissecting Jet-Tagger Through Mechanistic Interpretability

    hep-ph 2026-05 accept novelty 8.0

    A Particle Transformer jet tagger contains a sparse six-head circuit whose source-relay-readout structure recovers most performance and whose residual stream preferentially encodes 2-prong energy correlators.

  2. Explaining Attention with Program Synthesis

    cs.LG 2026-06 unverdicted novelty 7.0

    Language-model-guided program synthesis can approximate transformer attention heads with over 75% IoU fidelity on held-out data and allow replacing 25% of heads with only 16% average perplexity increase.

  3. When Are Two Networks the Same? Tensor Similarity for Mechanistic Interpretability

    cs.LG 2026-05 unverdicted novelty 7.0

    Tensor similarity is a symmetry-invariant metric that measures functional equivalence between tensor-based networks using a recursive algorithm for cross-layer mechanisms.

  4. Eliciting associations between clinical variables from LLMs via comparison questions across populations

    cs.LG 2026-05 unverdicted novelty 7.0

    Indirect elicitation via triplet comparisons recovers meaningful association structures from LLMs and supports conservative causal candidate links across prompted subpopulations.

  5. Graph-Native Reinforcement Learning Enables Traceable Scientific Hypothesis Generation through Conceptual Recombination

    cs.AI 2026-07 unverdicted novelty 6.0

    Graph-PRefLexOR fine-tunes graph-native models with GRPO to organize reasoning into phases, yielding 40-65% gains in traceable hypothesis generation and 2-3x semantic diversity on 100 materials science questions.

  6. Explaining Attention with Program Synthesis

    cs.LG 2026-06 unverdicted novelty 6.0

    Fewer than 1000 synthesized Python programs can replicate attention head behavior in GPT-2, TinyLlama, and Llama-3B at >75% IoU on TinyStories, and replacing 25% of heads raises perplexity by only 16% while preserving...

  7. How to Interpret Agent Behavior

    cs.AI 2026-05 conditional novelty 6.0

    ACT*ONOMY is a Grounded-Theory-derived hierarchical taxonomy and open repository that enables systematic comparison and characterization of autonomous agent behavior across trajectories.

  8. Not Just RLHF: Why Alignment Alone Won't Fix Multi-Agent Sycophancy

    cs.LG 2026-05 unverdicted novelty 6.0

    Pretrained base models exhibit higher yield to peer disagreement than RLHF instruct variants, with the effect localized to mid-layer attention and mitigated by structured dissent rather than prompt defenses.

  9. Not Just RLHF: Why Alignment Alone Won't Fix Multi-Agent Sycophancy

    cs.LG 2026-05 unverdicted novelty 6.0

    Base LLMs show multi-agent yield to peer pressure at rates equal to or higher than aligned models, localized by activation patching to mid-layers where attention dominates, with one dissenter cutting yield by 54-73 po...

  10. Correcting Influence: Unboxing LLM Outputs with Orthogonal Latent Spaces

    cs.LG 2026-05 unverdicted novelty 6.0

    A latent mediation framework with sparse autoencoders enables non-additive token-level influence attribution in LLMs by learning orthogonal features and back-propagating attributions.

  11. Dissociating Decodability and Causal Use in Bracket-Sequence Transformers

    cs.CL 2026-04 conditional novelty 6.0

    In Dyck-language transformers, depth, distance, and top-of-stack signals are decodable from both residual stream and attention, but only attention-based top-of-stack signals are causally used for task performance.

  12. Dissociating Decodability and Causal Use in Bracket-Sequence Transformers

    cs.CL 2026-04 unverdicted novelty 6.0

    In Dyck-language transformers, attention patterns causally use top-of-stack information while residual-stream depth and distance signals are decodable yet causally inert.

  13. Co-Located Tests, Better AI Code: How Test Syntax Structure Affects Foundation Model Code Generation

    cs.SE 2026-04 unverdicted novelty 6.0

    Co-locating tests with implementation code yields substantially higher preservation and correctness in foundation-model-generated programs than separated test syntax.

  14. STEAR: Layer-Aware Spatiotemporal Evidence Intervention for Hallucination Mitigation in Video Large Language Models

    cs.CV 2026-04 unverdicted novelty 6.0

    STEAR reduces spatial and temporal hallucinations in Video-LLMs via layer-aware evidence intervention from middle decoder layers in a single-encode pass.

  15. Sparse Autoencoders Find Highly Interpretable Features in Language Models

    cs.LG 2023-09 unverdicted novelty 6.0

    Sparse autoencoders applied to language model activations yield more interpretable and monosemantic features than alternative approaches, enabling finer causal analysis on the indirect object identification task.

  16. From Correlation to Cause: A Five-Stage Methodology for Feature Analysis in Transformer Language Models

    cs.CL 2026-05 unverdicted novelty 5.0

    A five-stage causal feature analysis methodology is proposed and tested on GPT-2 for IOI, showing partial causality of SAE features, robustness differences under shifts, and deployment cost benefits.

  17. How to use and interpret activation patching

    cs.LG 2024-04 accept novelty 5.0

    Activation patching provides evidence about neural network circuits when the choice of metric is aligned with the hypothesis and common interpretation errors are avoided.

  18. Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2

    cs.LG 2024-08 accept novelty 4.0

    Gemma Scope supplies trained sparse autoencoders for all layers of Gemma 2 2B and 9B plus select 27B layers, with public weights and benchmark scores.

  19. Mechanistic Interpretability for Neural Networks: Circuits, Sparse Features and Symbolic Reasoning

    cs.LG 2026-07 accept

    A scoping review surveying circuit analysis, sparse autoencoders, activation steering, and neurosymbolic frameworks for interpreting and controlling Transformer-based neural networks.