{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ON5CM22ZEQCAYCTS3T4P2EQ6Y7","short_pith_number":"pith:ON5CM22Z","schema_version":"1.0","canonical_sha256":"737a266b5924040c0a72dcf8fd121ec7d7ba0ec3b90be8a9882d941ce25bca30","source":{"kind":"arxiv","id":"2607.14111","version":1},"attestation_state":"computed","paper":{"title":"Introspection Fine-Tuning (IFT): Training Small LLMs to Introspect","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ely Hahami, Ishaan Sinha, Lavik Jain","submitted_at":"2026-05-08T05:30:26Z","abstract_excerpt":"Can small language models detect and report on perturbations their own internal activations? We investigate this question through the lens of activation steering: injecting concept vectors into a model's residual stream and measuring whether the model can accurately report on the perturbation. We first show that the binary detection paradigm used in prior work -- prompting the model to answer Yes'' or No'' to whether it detects an injected thought -- is confounded in small models, as steering biases the model toward affirmative responses regardless of the question content. We therefore propose"},"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":"2607.14111","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-08T05:30:26Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"29c640560aefc347786cffb16a9b7a074f1a28b8275ac4132c3046cfc21b126c","abstract_canon_sha256":"29acafd3109c852f485a56586599115dba013610c38f2c538e708e282c26eae4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-17T00:20:52.533733Z","signature_b64":"T30t1EIdcc+MGZr8mNTNyb6TH8e7hcyc8jrj7UUvShz+uRKrVGhK8i+HaPyIHywxIEZ/y3ASXteVqb577Fj0Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"737a266b5924040c0a72dcf8fd121ec7d7ba0ec3b90be8a9882d941ce25bca30","last_reissued_at":"2026-07-17T00:20:52.532925Z","signature_status":"signed_v1","first_computed_at":"2026-07-17T00:20:52.532925Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Introspection Fine-Tuning (IFT): Training Small LLMs to Introspect","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ely Hahami, Ishaan Sinha, Lavik Jain","submitted_at":"2026-05-08T05:30:26Z","abstract_excerpt":"Can small language models detect and report on perturbations their own internal activations? We investigate this question through the lens of activation steering: injecting concept vectors into a model's residual stream and measuring whether the model can accurately report on the perturbation. We first show that the binary detection paradigm used in prior work -- prompting the model to answer Yes'' or No'' to whether it detects an injected thought -- is confounded in small models, as steering biases the model toward affirmative responses regardless of the question content. We therefore propose"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.14111","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/2607.14111/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":"2607.14111","created_at":"2026-07-17T00:20:52.533341+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.14111v1","created_at":"2026-07-17T00:20:52.533341+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.14111","created_at":"2026-07-17T00:20:52.533341+00:00"},{"alias_kind":"pith_short_12","alias_value":"ON5CM22ZEQCA","created_at":"2026-07-17T00:20:52.533341+00:00"},{"alias_kind":"pith_short_16","alias_value":"ON5CM22ZEQCAYCTS","created_at":"2026-07-17T00:20:52.533341+00:00"},{"alias_kind":"pith_short_8","alias_value":"ON5CM22Z","created_at":"2026-07-17T00:20:52.533341+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/ON5CM22ZEQCAYCTS3T4P2EQ6Y7","json":"https://pith.science/pith/ON5CM22ZEQCAYCTS3T4P2EQ6Y7.json","graph_json":"https://pith.science/api/pith-number/ON5CM22ZEQCAYCTS3T4P2EQ6Y7/graph.json","events_json":"https://pith.science/api/pith-number/ON5CM22ZEQCAYCTS3T4P2EQ6Y7/events.json","paper":"https://pith.science/paper/ON5CM22Z"},"agent_actions":{"view_html":"https://pith.science/pith/ON5CM22ZEQCAYCTS3T4P2EQ6Y7","download_json":"https://pith.science/pith/ON5CM22ZEQCAYCTS3T4P2EQ6Y7.json","view_paper":"https://pith.science/paper/ON5CM22Z","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.14111&json=true","fetch_graph":"https://pith.science/api/pith-number/ON5CM22ZEQCAYCTS3T4P2EQ6Y7/graph.json","fetch_events":"https://pith.science/api/pith-number/ON5CM22ZEQCAYCTS3T4P2EQ6Y7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ON5CM22ZEQCAYCTS3T4P2EQ6Y7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ON5CM22ZEQCAYCTS3T4P2EQ6Y7/action/storage_attestation","attest_author":"https://pith.science/pith/ON5CM22ZEQCAYCTS3T4P2EQ6Y7/action/author_attestation","sign_citation":"https://pith.science/pith/ON5CM22ZEQCAYCTS3T4P2EQ6Y7/action/citation_signature","submit_replication":"https://pith.science/pith/ON5CM22ZEQCAYCTS3T4P2EQ6Y7/action/replication_record"}},"created_at":"2026-07-17T00:20:52.533341+00:00","updated_at":"2026-07-17T00:20:52.533341+00:00"}