Papers › Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small

Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small

1 Nov 2022arXiv:2211.00593archive 2025-07-28

Kevin Wang, Alexandre Variengien, Arthur Conmy, Buck Shlegeris, Jacob Steinhardt

Research in mechanistic interpretability seeks to explain behaviors of machine learning models in terms of their internal components. However, most previous work either focuses on simple behaviors in small models, or describes complicated behaviors in larger models with broad strokes. In this work, we bridge this gap by presenting an explanation for how GPT-2 small performs a natural language task called indirect object identification (IOI). Our explanation encompasses 26 attention heads grouped into 7 main classes, which we discovered using a combination of interpretability approaches relying on causal interventions. To our knowledge, this investigation is the largest end-to-end attempt at reverse-engineering a natural behavior "in the wild" in a language model. We evaluate the reliability of our explanation using three quantitative criteria--faithfulness, completeness and minimality. Though these criteria support our explanation, they also point to remaining gaps in our understanding. Our work provides evidence that a mechanistic understanding of large ML models is feasible, opening opportunities to scale our understanding to both larger models and more complex tasks.

PaperPDFCodeCode Syntology ran

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

No sample from the archive's official repository (redwoodresearch/easy-transformer) is linked to this paper; the 13 samples below come from 1 community repository. Syntology harvested code from redwoodresearch/easy-transformer for another paper.

redwoodresearch/easy-transformer: Syntology's graph links these samples to Towards Best Practices of Activation Patching in Language Models: Metrics and Methods (arXiv:2309.16042; that paper's own run record: 5 ran (of which 0 constructed an object rather than computing a result; 5 with no instrument failure: 0 honoured, 0 violated, 5 with no contract checked; 0 where Syntology's instrument failed) · 1 unverified); nothing checks that they implement this paper's method.

Syntology 13 code samples harvested from 1 repository linked to this paper: 9 ran (of which 0 constructed an object rather than computing a result; 9 with no instrument failure: 0 honoured, 0 violated, 9 with no contract checked; 0 where Syntology's instrument failed) · 4 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code; here there are none.

By repository: community (archive-listed): 13 samples from 1 repository, 9 ran (9 with no instrument failure; 0 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

No sample is linked to this paper from 5 of its repositories that Syntology harvested for other papers:

Levels written on 2026-09-25 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

redwoodresearch/easy-transformer officialmentioned in paperpytorchMITSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Towards Best Practices of Activation Patching in Language Models: Metrics and Methods (arXiv:2309.16042; that paper's own run record: 5 ran (of which 0 constructed an object rather than computing a result; 5 with no instrument failure: 0 honoured, 0 violated, 5 with no contract checked; 0 where Syntology's instrument failed) · 1 unverified); nothing checks that they implement this paper's method. report
FlyingPumba/circuits-benchmark mentioned on GitHubjaxMITSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to InterpBench: Semi-Synthetic Transformers for Evaluating Mechanistic Interpretability Techniques (arXiv:2407.14494; that paper's own run record: 3 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 3 where Syntology's instrument failed) · 1 unverified); nothing checks that they implement this paper's method. report
callummcdougall/arena_3.0 mentioned on GitHubjaxSyntology: not harvested report
flyingpumba/interpbench mentioned on GitHubjaxMITSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to InterpBench: Semi-Synthetic Transformers for Evaluating Mechanistic Interpretability Techniques (arXiv:2407.14494; that paper's own run record: 3 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 3 where Syntology's instrument failed) · 1 unverified); nothing checks that they implement this paper's method. report
openai/transformer-debugger mentioned on GitHubpytorchMITSyntology: 13 samples, 9 ran (0 instrument-failed) · 4 unverified report
openmoss/language-model-saes mentioned on GitHubjaxSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Llama Scope: Extracting Millions of Features from Llama-3.1-8B with Sparse Autoencoders (arXiv:2410.20526; that paper's own run record: 4 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 0 honoured, 0 violated, 3 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method. report
technion-cs-nlp/peap mentioned on GitHubjaxMITSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Position-aware Automatic Circuit Discovery (arXiv:2502.04577; that paper's own run record: 7 ran (of which 0 constructed an object rather than computing a result; 7 with no instrument failure: 0 honoured, 0 violated, 7 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method. report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.

Code Syntology ran Syntology

13 samples harvested: 9 ran (of which 0 constructed an object rather than computing a result; 9 with no instrument failure: 0 honoured, 0 violated, 9 with no contract checked; 0 where Syntology's instrument failed) · 4 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page has not changed since , the first build that kept a record date for it; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

9ran
4unverified

Licence: 0 of the 13 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from openai/transformer-debugger. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity.

openai/transformer-debugger community (archive-listed) 13 samples, 9 ran (0 instrument-failed) · 4 unverified
async_exponential_backoff neuron_explainer/api_client.py ran MIT (permissive) · 37ed7c090dd47a16 · report fixture: signature retry_on:callable/0/? · returned function · level recorded 2026-09-25
causal_attn_mask neuron_explainer/models/transformer.py ran MIT (permissive) · 7a5b341cfbfce5c4 · report fixture: signature size:int/0/int, device:str/0/str · returned Tensor · level recorded 2026-09-25
file_exists in 8 papers neuron_explainer/file_utils.py ran MIT (permissive) · 3bf23c5e5bb20c36 · report fixture: signature filepath:str/0/str · returned bool · level recorded 2026-09-25
grad_hook_wrapper neuron_explainer/models/hooks.py ran MIT (permissive) · cbf09830a05cc1ec · report fixture: signature grad_hook:callable/0/? · returned function · level recorded 2026-09-25
is_api_error neuron_explainer/api_client.py ran MIT (permissive) · 0d9b56dd70f4d1e7 · report fixture: signature err:HTTPStatusError/0/? · returned bool · level recorded 2026-09-25
read_single_async neuron_explainer/file_utils.py ran MIT (permissive) · 5e57cfebe2fb8845 · report fixture: signature filepath:str/0/str · returned coroutine · level recorded 2026-09-25
safe_div in 29 papers neuron_explainer/models/transformer.py ran fingerprinted MIT (permissive) · 16add703a8e2471b · report fixture: signature numerator:int/0/int, denominator:int/0/int · returned int · level recorded 2026-09-25
split_heads in 10 papers neuron_explainer/models/transformer.py ran fingerprinted MIT (permissive) · 3ef93b73d7c3c9ea · report fixture: signature Z:Tensor/3/float, n_heads:int/0/int · returned Tensor · level recorded 2026-09-25
to_camel neuron_explainer/pydantic/camel_case_base_model.py ran fingerprinted MIT (permissive) · 286523200b9c09a1 · report fixture: signature string:str/0/str · returned str · level recorded 2026-09-25
get_corresponding_object neuron_explainer/activation_server/interactive_model.py unverified MIT (permissive) · 35ca65faef55b637 · report
get_dimension_index_of_weight_location_type neuron_explainer/models/model_component_registry.py unverified MIT (permissive) · 817bdde39fefce9d · report
get_standard_model_spec neuron_explainer/models/model_registry.py unverified MIT (permissive) · be193c0532e4250d · report
list_autoencoder_names neuron_explainer/models/model_registry.py unverified MIT (permissive) · 0bafca70b42d4ed7 · report

Tasks

Language ModelingLanguage Modelling

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Methods

AdamAttentionAttention DropoutBPECosine AnnealingDense ConnectionsDiscriminative Fine-TuningDropoutGPT-2Layer NormalizationLinear LayerLinear Warmup With Cosine AnnealingMulti-Head AttentionResidual ConnectionSoftmaxWeight Decay

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections