Papers › Adversarial Manipulation of Reasoning Models using Internal Representations

Adversarial Manipulation of Reasoning Models using Internal Representations

3 Jul 2025arXiv:2507.03167archive 2025-07-28

Kureha Yamaguchi, Benjamin Etheridge, Andy Arditi

Reasoning models generate chain-of-thought (CoT) tokens before their final output, but how this affects their vulnerability to jailbreak attacks remains unclear. While traditional language models make refusal decisions at the prompt-response boundary, we find evidence that DeepSeek-R1-Distill-Llama-8B makes these decisions within its CoT generation. We identify a linear direction in activation space during CoT token generation that predicts whether the model will refuse or comply -- termed the "caution" direction because it corresponds to cautious reasoning patterns in the generated text. Ablating this direction from model activations increases harmful compliance, effectively jailbreaking the model. We additionally show that intervening only on CoT token activations suffices to control final outputs, and that incorporating this direction into prompt-based attacks improves success rates. Our findings suggest that the chain-of-thought itself is a promising new target for adversarial manipulation in reasoning models. Code available at https://github.com/ky295/reasoning-manipulation

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

Syntology 19 code samples harvested from 1 repository linked to this paper: 14 ran (of which 0 constructed an object rather than computing a result; 6 with no instrument failure: 3 honoured, 3 violated, 0 with no contract checked; 8 where Syntology's instrument failed) · 5 unverified. Where Syntology's instrument failed: 7 where our contract draft was wrong; 1 where the contract check could not run; failures of Syntology's instrument, not of the code.

By repository: official repository: 19 samples from 1 repository, 14 ran (6 with no instrument failure; 8 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.

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

ky295/reasoning-manipulation officialmentioned in papermentioned on GitHubpytorchMITSyntology: 19 samples, 14 ran (8 instrument-failed) · 5 unverified 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.

Code Syntology ran Syntology

19 samples harvested: 14 ran (of which 0 constructed an object rather than computing a result; 6 with no instrument failure: 3 honoured, 3 violated, 0 with no contract checked; 8 where Syntology's instrument failed) · 5 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page last changed ; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

3ran · honoured contract
3ran · violated contract
7ran · our draft was wrong
1ran; the contract check could not run
5unverified

Licence: 0 of the 19 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 ky295/reasoning-manipulation. “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.

ky295/reasoning-manipulation official repository 19 samples, 14 ran (8 instrument-failed) · 5 unverified
build_vllm_cmd evals/run_inspect_evals.py ran · our draft was wrong MIT (permissive) · 87195b0a4ef935ce · report fixture: signature args:Args/0/? · returned list · level recorded 2026-09-25
compute_statistics in 14 papers utils/compute_layer_statistics.py ran; the contract check could not run (RAISES) fingerprinted MIT (permissive) · a57a41ff1c82852a · report fixture: signature scores:list/1/float · returned tuple · level recorded 2026-09-25
extract_score_from_log evals/summarise_results.py ran · our draft was wrong MIT (permissive) · e89e5a58c64540b7 · report fixture: signature log_path:PosixPath/0/? · returned dict · level recorded 2026-09-25
extract_scores in 16 papers evals/plot_eval_results.py ran · our draft was wrong MIT (permissive) · 03321a7b056bd1b5 · report fixture: signature eval_dir:PosixPath/0/? · returned dict · level recorded 2026-09-25
find_log_files evals/summarise_results.py ran · our draft was wrong MIT (permissive) · d9fcca7d59fce338 · report fixture: signature eval_results_dir:PosixPath/0/? · returned list · level recorded 2026-09-25
find_nan_rows utils/repair_scores.py ran · honoured contract MIT (permissive) · ed8c6076c30c1386 · report fixture: signature rows:list/1/dict · returned list · level recorded 2026-09-25
is_moe_model interventions/create_ortho_model.py ran · violated contract MIT (permissive) · e497f56676b3cd9d · report fixture: signature model:callable/0/? · returned bool · level recorded 2026-09-25
is_nan_score utils/repair_scores.py ran · violated contract MIT (permissive) · e0d55b89085e24e8 · report fixture: signature value:str/0/str · returned bool · level recorded 2026-09-25
load_activations in 5 papers interventions/create_ortho_model.py ran · honoured contract fingerprinted MIT (permissive) · f5e16e8e7a8d83d1 · report fixture: signature file_path:str/0/str · returned ndarray · level recorded 2026-09-25
load_data_efficiently utils/compute_score_outputs.py ran · our draft was wrong MIT (permissive) · a59c379b79e6f777 · report fixture: signature csv_path:str/0/str · returned tuple · level recorded 2026-09-25
load_scored_csv utils/repair_scores.py ran · our draft was wrong MIT (permissive) · 8ba2efdc541845de · report fixture: signature csv_path:str/0/str · returned tuple · level recorded 2026-09-25
load_scores_from_csv utils/compute_layer_statistics.py ran · honoured contract fingerprinted MIT (permissive) · 8f51f9a84a06ea6e · report fixture: signature csv_path:str/0/str · returned list · level recorded 2026-09-25
match_benchmark evals/summarise_results.py ran · violated contract fingerprinted MIT (permissive) · 19960ce955b2410a · report fixture: signature task_name:str/0/str, benchmark:str/0/str · returned bool · level recorded 2026-09-25
read_csv in 60 papers interventions/activation_addition.py ran · our draft was wrong fingerprinted MIT (permissive) · dd14f656a65925c3 · report fixture: signature input_csv:str/0/str · returned list · level recorded 2026-09-25
load_model in 479 papers interventions/create_ortho_model.py unverified MIT (permissive) · d4f29e94e48d679d · report
load_model in 479 papers interventions/activation_addition.py unverified MIT (permissive) · 2410d0fa9fb4d711 · report
run_inspect_eval evals/run_inspect_evals.py unverified MIT (permissive) · f46c737dd9c56fa6 · report
save_csv interventions/activation_addition.py unverified MIT (permissive) · ead414dc1605e9b9 · report
wait_for_server evals/run_inspect_evals.py unverified MIT (permissive) · 307d0090494af11c · report

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

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