{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:TQLEP4SMFW35LYST3FQ55SVWDC","short_pith_number":"pith:TQLEP4SM","schema_version":"1.0","canonical_sha256":"9c1647f24c2db7d5e253d961decab6188bccf037eda8360a1962ed63b8954420","source":{"kind":"arxiv","id":"2606.00726","version":1},"attestation_state":"computed","paper":{"title":"Latent Reward Steering: An Adaptive Inference-Time Framework that Implicitly Promotes Cognitive Behaviors in Reasoning LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Can Jin, Chenxi Huang, Dexu Yu, Dimitris N. Metaxas, Guanyu Zhu, Hongwu Peng, Jiakang Li, Ronghao Chen, Xuanqi Lan, Yang Zhou, Youhua Li","submitted_at":"2026-05-30T13:38:06Z","abstract_excerpt":"Strong reasoning depends not only on model knowledge but also on how effectively cognitive behaviors are deployed during generation. Existing methods often rely on explicit behavior-level control, making them insufficiently adaptive when failures and required corrections vary across reasoning states, tasks, and models. To this end, we propose Latent Reward Steering (LRS), an adaptive inference-time framework that promotes cognitive behaviors by optimizing the sparse-autoencoder (SAE) latent states that implicitly carry them. Rather than relying on predefined cognitive behaviors or steering dir"},"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":"2606.00726","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-05-30T13:38:06Z","cross_cats_sorted":[],"title_canon_sha256":"84841c0c41c86532509dfbe2a1129583e4df4d1198c3a74a1ba57f960a3aa2a2","abstract_canon_sha256":"8bcffb2121093e1e8f8665448b896c5f380db15a65712220334d20bc50615343"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T01:04:03.958847Z","signature_b64":"e2HwDFV+KZ7rv4S6IhKrxIxCsJpZLPjwsj+xmaed0i7WuJ7TbWkahjWk8PRDFD4l//yVyto0XjSV1cE5of/HBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9c1647f24c2db7d5e253d961decab6188bccf037eda8360a1962ed63b8954420","last_reissued_at":"2026-06-02T01:04:03.958443Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T01:04:03.958443Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Latent Reward Steering: An Adaptive Inference-Time Framework that Implicitly Promotes Cognitive Behaviors in Reasoning LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Can Jin, Chenxi Huang, Dexu Yu, Dimitris N. Metaxas, Guanyu Zhu, Hongwu Peng, Jiakang Li, Ronghao Chen, Xuanqi Lan, Yang Zhou, Youhua Li","submitted_at":"2026-05-30T13:38:06Z","abstract_excerpt":"Strong reasoning depends not only on model knowledge but also on how effectively cognitive behaviors are deployed during generation. Existing methods often rely on explicit behavior-level control, making them insufficiently adaptive when failures and required corrections vary across reasoning states, tasks, and models. To this end, we propose Latent Reward Steering (LRS), an adaptive inference-time framework that promotes cognitive behaviors by optimizing the sparse-autoencoder (SAE) latent states that implicitly carry them. Rather than relying on predefined cognitive behaviors or steering dir"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.00726","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/2606.00726/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":"2606.00726","created_at":"2026-06-02T01:04:03.958504+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.00726v1","created_at":"2026-06-02T01:04:03.958504+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.00726","created_at":"2026-06-02T01:04:03.958504+00:00"},{"alias_kind":"pith_short_12","alias_value":"TQLEP4SMFW35","created_at":"2026-06-02T01:04:03.958504+00:00"},{"alias_kind":"pith_short_16","alias_value":"TQLEP4SMFW35LYST","created_at":"2026-06-02T01:04:03.958504+00:00"},{"alias_kind":"pith_short_8","alias_value":"TQLEP4SM","created_at":"2026-06-02T01:04:03.958504+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/TQLEP4SMFW35LYST3FQ55SVWDC","json":"https://pith.science/pith/TQLEP4SMFW35LYST3FQ55SVWDC.json","graph_json":"https://pith.science/api/pith-number/TQLEP4SMFW35LYST3FQ55SVWDC/graph.json","events_json":"https://pith.science/api/pith-number/TQLEP4SMFW35LYST3FQ55SVWDC/events.json","paper":"https://pith.science/paper/TQLEP4SM"},"agent_actions":{"view_html":"https://pith.science/pith/TQLEP4SMFW35LYST3FQ55SVWDC","download_json":"https://pith.science/pith/TQLEP4SMFW35LYST3FQ55SVWDC.json","view_paper":"https://pith.science/paper/TQLEP4SM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.00726&json=true","fetch_graph":"https://pith.science/api/pith-number/TQLEP4SMFW35LYST3FQ55SVWDC/graph.json","fetch_events":"https://pith.science/api/pith-number/TQLEP4SMFW35LYST3FQ55SVWDC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TQLEP4SMFW35LYST3FQ55SVWDC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TQLEP4SMFW35LYST3FQ55SVWDC/action/storage_attestation","attest_author":"https://pith.science/pith/TQLEP4SMFW35LYST3FQ55SVWDC/action/author_attestation","sign_citation":"https://pith.science/pith/TQLEP4SMFW35LYST3FQ55SVWDC/action/citation_signature","submit_replication":"https://pith.science/pith/TQLEP4SMFW35LYST3FQ55SVWDC/action/replication_record"}},"created_at":"2026-06-02T01:04:03.958504+00:00","updated_at":"2026-06-02T01:04:03.958504+00:00"}