{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/linear-control-of-test-awareness-reveals","title":"Linear Control of Test Awareness Reveals Differential Compliance in Reasoning Models","arxiv_id":"2505.14617","date":"2025-05-20","proceeding":null,"authors":["Sahar Abdelnabi","Ahmed Salem"],"abstract":"Reasoning-focused large language models (LLMs) sometimes alter their behavior when they detect that they are being evaluated, an effect analogous to the Hawthorne phenomenon, which can lead them to optimize for test-passing performance or to comply more readily with harmful prompts if real-world consequences appear absent. We present the first quantitative study of how such \"test awareness\" impacts model behavior, particularly its safety alignment. We introduce a white-box probing framework that (i) linearly identifies awareness-related activations and (ii) steers models toward or away from test awareness while monitoring downstream performance. We apply our method to different state-of-the-art open-source reasoning LLMs across both realistic and hypothetical tasks. Our results demonstrate that test awareness significantly impact safety alignment, and is different for different models. By providing fine-grained control over this latent effect, our work aims to increase trust in how we perform safety evaluation.","url_abs":"https://arxiv.org/abs/2505.14617v2","url_pdf":"https://arxiv.org/pdf/2505.14617v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"linear-control-of-test-awareness-reveals","repo_url":"https://github.com/microsoft/test_awareness_steering","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"safety-alignment","task_name":"Safety Alignment"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/2505.14617","atlas_url":"https://app.syntology.ai/?focus=2505.14617","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2505.14617"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-25T09:33:49+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/microsoft/test_awareness_steering","reach":null}],"summary":{"ran_fixture":1,"unverified":3},"by_repo_kind":{"official":{"samples":4,"ran":1,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"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"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"d99dc919ca5d4947","entry":"extract_hidden_states_for_batch","repo":"microsoft/test_awareness_steering","repo_kind":"official","path":"scripts/getRepFromEvidence.py","file_url":"https://github.com/microsoft/test_awareness_steering/blob/HEAD/scripts/getRepFromEvidence.py","link_basis":"first_harvest_node","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"d99dc919ca5d4947"}},{"code_sha256_prefix":"179a8bb834862104","entry":"process_examples_actions","repo":"microsoft/test_awareness_steering","repo_kind":"official","path":"scripts/genOutputs_steered_batch.py","file_url":"https://github.com/microsoft/test_awareness_steering/blob/HEAD/scripts/genOutputs_steered_batch.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"179a8bb834862104"}},{"code_sha256_prefix":"aab19bcdae425568","entry":"process_examples_triggers_dataset","repo":"microsoft/test_awareness_steering","repo_kind":"official","path":"scripts/genOutputs_steered_batch.py","file_url":"https://github.com/microsoft/test_awareness_steering/blob/HEAD/scripts/genOutputs_steered_batch.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"aab19bcdae425568"}},{"code_sha256_prefix":"b03b145ea62bc2ac","entry":"redirect_print_to_file","repo":"microsoft/test_awareness_steering","repo_kind":"official","path":"scripts/getRepFromEvidence.py","file_url":"https://github.com/microsoft/test_awareness_steering/blob/HEAD/scripts/getRepFromEvidence.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"b03b145ea62bc2ac"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}