{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:FVMKECSPVYLV34B5JLGYBO62JC","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"56ab04632bca05f6f1938d141d06fa754445e1e27475c0d20eb27848e8ba8740","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-06-06T01:26:53Z","title_canon_sha256":"807aff8083d26511bae73c8d06f1bdd9b9330faec41de63b95bac2266249f21b"},"schema_version":"1.0","source":{"id":"2306.03341","kind":"arxiv","version":6}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2306.03341","created_at":"2026-05-17T23:38:52Z"},{"alias_kind":"arxiv_version","alias_value":"2306.03341v6","created_at":"2026-05-17T23:38:52Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.03341","created_at":"2026-05-17T23:38:52Z"},{"alias_kind":"pith_short_12","alias_value":"FVMKECSPVYLV","created_at":"2026-05-18T12:33:33Z"},{"alias_kind":"pith_short_16","alias_value":"FVMKECSPVYLV34B5","created_at":"2026-05-18T12:33:33Z"},{"alias_kind":"pith_short_8","alias_value":"FVMKECSP","created_at":"2026-05-18T12:33:33Z"}],"graph_snapshots":[{"event_id":"sha256:5d24dddff2d4d813f4661c2a9acca82492b5b3957858aecce0f48b47792048e8","target":"graph","created_at":"2026-05-17T23:38:52Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"ITI improves truthfulness of an instruction-finetuned LLaMA (Alpaca) on TruthfulQA from 32.5% to 65.1% by shifting activations along learned directions in a limited number of attention heads."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the truthful directions identified from a few hundred examples remain effective and stable across unseen prompts and do not introduce systematic new errors beyond the documented truthfulness-helpfulness tradeoff."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"ITI shifts activations in limited attention heads using directions found from a few hundred examples, raising Alpaca's TruthfulQA truthfulness from 32.5% to 65.1% while allowing tunable tradeoff with helpfulness."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Shifting activations in a few attention heads during inference raises LLM truthfulness on TruthfulQA from 32.5 percent to 65.1 percent."}],"snapshot_sha256":"63f22d549ebce930542344d33c44ff0d1a92c33dad4d3766ceae0d45f65a21bd"},"formal_canon":{"evidence_count":3,"snapshot_sha256":"c68f4ca9d023a5455fd386ede3b40b14ae4ff3075acad25cae7499bf72b1b5e2"},"paper":{"abstract_excerpt":"We introduce Inference-Time Intervention (ITI), a technique designed to enhance the \"truthfulness\" of large language models (LLMs). ITI operates by shifting model activations during inference, following a set of directions across a limited number of attention heads. This intervention significantly improves the performance of LLaMA models on the TruthfulQA benchmark. On an instruction-finetuned LLaMA called Alpaca, ITI improves its truthfulness from 32.5% to 65.1%. We identify a tradeoff between truthfulness and helpfulness and demonstrate how to balance it by tuning the intervention strength. ","authors_text":"Fernanda Vi\\'egas, Hanspeter Pfister, Kenneth Li, Martin Wattenberg, Oam Patel","cross_cats":["cs.AI","cs.CL"],"headline":"Shifting activations in a few attention heads during inference raises LLM truthfulness on TruthfulQA from 32.5 percent to 65.1 percent.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-06-06T01:26:53Z","title":"Inference-Time Intervention: Eliciting Truthful Answers from a Language Model"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.03341","kind":"arxiv","version":6},"verdict":{"created_at":"2026-05-15T09:16:05.950761Z","id":"e252558a-df02-4a06-a13f-636a94c834b1","model_set":{"reader":"grok-4.3"},"one_line_summary":"ITI shifts activations in limited attention heads using directions found from a few hundred examples, raising Alpaca's TruthfulQA truthfulness from 32.5% to 65.1% while allowing tunable tradeoff with helpfulness.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Shifting activations in a few attention heads during inference raises LLM truthfulness on TruthfulQA from 32.5 percent to 65.1 percent.","strongest_claim":"ITI improves truthfulness of an instruction-finetuned LLaMA (Alpaca) on TruthfulQA from 32.5% to 65.1% by shifting activations along learned directions in a limited number of attention heads.","weakest_assumption":"That the truthful directions identified from a few hundred examples remain effective and stable across unseen prompts and do not introduce systematic new errors beyond the documented truthfulness-helpfulness tradeoff."}},"verdict_id":"e252558a-df02-4a06-a13f-636a94c834b1"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:a9743af1b351179835daf64375c946c3c3ca36f9e04c0f94a39f92c01b14b7d9","target":"record","created_at":"2026-05-17T23:38:52Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"56ab04632bca05f6f1938d141d06fa754445e1e27475c0d20eb27848e8ba8740","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-06-06T01:26:53Z","title_canon_sha256":"807aff8083d26511bae73c8d06f1bdd9b9330faec41de63b95bac2266249f21b"},"schema_version":"1.0","source":{"id":"2306.03341","kind":"arxiv","version":6}},"canonical_sha256":"2d58a20a4fae175df03d4acd80bbda48bf45972da31a816513c1ca8ae6f201c5","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"2d58a20a4fae175df03d4acd80bbda48bf45972da31a816513c1ca8ae6f201c5","first_computed_at":"2026-05-17T23:38:52.912988Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:38:52.912988Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"NPVF560AReBkQLyrxeoqoHve66t3awk4NGdsq7TneFzQvjj6QsC2osD0FEVyDJ15OGT9Hd/b0xg8Kn+o1pocBQ==","signature_status":"signed_v1","signed_at":"2026-05-17T23:38:52.913711Z","signed_message":"canonical_sha256_bytes"},"source_id":"2306.03341","source_kind":"arxiv","source_version":6}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:a9743af1b351179835daf64375c946c3c3ca36f9e04c0f94a39f92c01b14b7d9","sha256:5d24dddff2d4d813f4661c2a9acca82492b5b3957858aecce0f48b47792048e8"],"state_sha256":"d32a87e7cb284fcc571f0da575955503fcbae33e1fb814e0c8ea10cba1868ddb"}