SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2604.07615 · 2026

ADAG: Automatically Describing Attribution Graphs

Jacob Steinhardt, Sarah Schwettmann, Zhengxuan Wu, Aryaman Arora

arXiv · PDF · Open in the Atlas

Code that ran

5 of 8 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 1 violated it and 0 have no contract verdict; 4 failures of our instrument; 3 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2604.07615.json · badge: /api/ran/2604.07615.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
TransluceAI/circuits canonical (classifier label)
classifier basis: census
5 ran · 3 no recorded run
FunctionStatusWhere it lives
compute_auc Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
TransluceAI/circuits/circuits/analysis/score_features.py
pointer only (licence: NONE) · get_code("597e394b194e3552")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, TransluceAI/circuits when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

df_sum_over_tokens Ran
ran; our contract draft was wrong, not the code
TransluceAI/circuits/circuits/analysis/cluster.py
pointer only (licence: NONE) · get_code("883fb2e44e321089")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, TransluceAI/circuits when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

prepare_circuit_data Ran
ran; our contract draft was wrong, not the code
TransluceAI/circuits/circuits/analysis/cluster.py
pointer only (licence: NONE) · get_code("6ced0d04a85b01e0")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, TransluceAI/circuits when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

revert_stop_nonlinear_grad_for_llama Ran
ran; our contract draft was wrong, not the code
TransluceAI/circuits/circuits/core/grad.py
pointer only (licence: NONE) · get_code("c7f0b186aacc129a")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, TransluceAI/circuits when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

unit_norm Ran
ran, violated the contract we drafted
TransluceAI/circuits/circuits/analysis/cluster.py
pointer only (licence: NONE) · get_code("a9c4a50ada359104")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, TransluceAI/circuits when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

prepare_table_from_steering_results Not yet run
unverified (harvested, no recorded run)
TransluceAI/circuits/circuits/analysis/label.py
pointer only (licence: NONE) · get_code("8a3f6c1e2937e2e8")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, TransluceAI/circuits when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

shapley_attention_forward Not yet run
unverified (harvested, no recorded run)
TransluceAI/circuits/circuits/core/grad.py
pointer only (licence: NONE) · get_code("360a03c88395e631")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, TransluceAI/circuits when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

stop_nonlinear_grad_for_llama Not yet run
unverified (harvested, no recorded run)
TransluceAI/circuits/circuits/core/grad.py
pointer only (licence: NONE) · get_code("10067bc5d35b2be1")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, TransluceAI/circuits when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 8 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

In language model interpretability research, circuit tracing aims to identify which internal features causally contributed to a particular output and how they affected each other, with the goal of explaining the computations underlying some behaviour. However, all prior circuit tracing work has relied on ad-hoc human interpretation of the role that each feature in the circuit plays, via manual inspection of data artifacts such as the dataset examples the component activates on. We introduce ADAG, an end-to-end pipeline for describing these attribution graphs which is fully automated. To achieve this, we introduce attribution profiles which quantify the functional role of a feature via its input and output gradient effects. We then introduce a novel clustering algorithm for grouping features, and an LLM explainersimulator setup which generates and scores natural-language explanations of the functional role of these feature groups. We run our system on known human-analysed circuit-tracing tasks and recover interpretable circuits, and further show ADAG can find steerable clusters which are responsible for a harmful advice jailbreak in Llama 3.1 8B Instruct. github.com/TransluceAI/circuits ADAG: End-to-end circuit interpretation Circuit tracing Attribution profiles Output Collect gradient attribution for each feature. Input Filter for components with grad. attribution exceeding τ. Compute edge weights. Input attribution Output contribution Clustering Compute pertype, per-example similarities Aggregate similarities Spectral clustering on feature-vs.feature similarities cos.sim. Descriptions 0. Collect cluster attr. profiles input attr.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2604.07615")
get_code_for_paper("2604.07615")
have("2604.07615")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2604.07615.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2604.07615.svg)](https://syntology.ai/paper/2604.07615)

Connect an agent — have() is free.