{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:DNRKHSXG5G5NCCWQDARBC72DBH","short_pith_number":"pith:DNRKHSXG","schema_version":"1.0","canonical_sha256":"1b62a3cae6e9bad10ad01822117f4309c9d30c62d5e5b041c08213194c2c34d3","source":{"kind":"arxiv","id":"2606.29196","version":1},"attestation_state":"computed","paper":{"title":"Representational Depth of Evaluation Awareness Shifts With Scale in Open-Weight Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Archit Manek","submitted_at":"2026-06-28T04:48:17Z","abstract_excerpt":"Do language models know when they are being tested? This question matters for AI safety: a model that recognises an evaluation context could alter its behaviour strategically, making downstream benchmarks harder to interpret. Using 11 models spanning Qwen 2.5, Gemma 2, and Llama 3.2, we find a systematic size-dependent shift in representational depth: in both Qwen 2.5 and Gemma 2, the layer at which evaluation-awareness is most linearly recoverable moves from late layers in smaller models to early layers in larger ones. This suggests that scale changes not only the strength of evaluation-aware"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2606.29196","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-06-28T04:48:17Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"4d748175157c11f50f479b9e4936629a1f292054b1ccca8a0457a469e2833159","abstract_canon_sha256":"9ad71b1581099f522ab6965eb5193784bd14d997000023808806cc30e2e873b9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-30T01:17:56.784473Z","signature_b64":"mSPF9cJt3TgoQcUwUSyftR93Ld8zKs2ol70xrJkRC/wUK+vsagqmfANyzOrd+xPwAO5R9OAivTTNkdUG3cQ1Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1b62a3cae6e9bad10ad01822117f4309c9d30c62d5e5b041c08213194c2c34d3","last_reissued_at":"2026-06-30T01:17:56.783940Z","signature_status":"signed_v1","first_computed_at":"2026-06-30T01:17:56.783940Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Representational Depth of Evaluation Awareness Shifts With Scale in Open-Weight Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Archit Manek","submitted_at":"2026-06-28T04:48:17Z","abstract_excerpt":"Do language models know when they are being tested? This question matters for AI safety: a model that recognises an evaluation context could alter its behaviour strategically, making downstream benchmarks harder to interpret. Using 11 models spanning Qwen 2.5, Gemma 2, and Llama 3.2, we find a systematic size-dependent shift in representational depth: in both Qwen 2.5 and Gemma 2, the layer at which evaluation-awareness is most linearly recoverable moves from late layers in smaller models to early layers in larger ones. This suggests that scale changes not only the strength of evaluation-aware"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.29196","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2606.29196/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2606.29196","created_at":"2026-06-30T01:17:56.784020+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.29196v1","created_at":"2026-06-30T01:17:56.784020+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.29196","created_at":"2026-06-30T01:17:56.784020+00:00"},{"alias_kind":"pith_short_12","alias_value":"DNRKHSXG5G5N","created_at":"2026-06-30T01:17:56.784020+00:00"},{"alias_kind":"pith_short_16","alias_value":"DNRKHSXG5G5NCCWQ","created_at":"2026-06-30T01:17:56.784020+00:00"},{"alias_kind":"pith_short_8","alias_value":"DNRKHSXG","created_at":"2026-06-30T01:17:56.784020+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DNRKHSXG5G5NCCWQDARBC72DBH","json":"https://pith.science/pith/DNRKHSXG5G5NCCWQDARBC72DBH.json","graph_json":"https://pith.science/api/pith-number/DNRKHSXG5G5NCCWQDARBC72DBH/graph.json","events_json":"https://pith.science/api/pith-number/DNRKHSXG5G5NCCWQDARBC72DBH/events.json","paper":"https://pith.science/paper/DNRKHSXG"},"agent_actions":{"view_html":"https://pith.science/pith/DNRKHSXG5G5NCCWQDARBC72DBH","download_json":"https://pith.science/pith/DNRKHSXG5G5NCCWQDARBC72DBH.json","view_paper":"https://pith.science/paper/DNRKHSXG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.29196&json=true","fetch_graph":"https://pith.science/api/pith-number/DNRKHSXG5G5NCCWQDARBC72DBH/graph.json","fetch_events":"https://pith.science/api/pith-number/DNRKHSXG5G5NCCWQDARBC72DBH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DNRKHSXG5G5NCCWQDARBC72DBH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DNRKHSXG5G5NCCWQDARBC72DBH/action/storage_attestation","attest_author":"https://pith.science/pith/DNRKHSXG5G5NCCWQDARBC72DBH/action/author_attestation","sign_citation":"https://pith.science/pith/DNRKHSXG5G5NCCWQDARBC72DBH/action/citation_signature","submit_replication":"https://pith.science/pith/DNRKHSXG5G5NCCWQDARBC72DBH/action/replication_record"}},"created_at":"2026-06-30T01:17:56.784020+00:00","updated_at":"2026-06-30T01:17:56.784020+00:00"}