{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:BDYRVN4BIQ5SUXZXJD33R7UTIW","short_pith_number":"pith:BDYRVN4B","schema_version":"1.0","canonical_sha256":"08f11ab781443b2a5f3748f7b8fe93459e3d83bacaf31ea41fb83ee960891575","source":{"kind":"arxiv","id":"2601.03595","version":2},"attestation_state":"computed","paper":{"title":"Controllable LLM Reasoning via Sparse Autoencoder-Based Steering","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Boyi Deng, Dayiheng Liu, Fengli Xu, Fuli Feng, Mingfeng Xue, Wenjie Wang, Yi Fang","submitted_at":"2026-01-07T05:26:26Z","abstract_excerpt":"Large Reasoning Models (LRMs) exhibit human-like cognitive reasoning strategies (\\eg backtracking, cross-verification) during the reasoning process, which improves their performance on complex tasks. Currently, reasoning strategies are autonomously selected by LRMs themselves. However, such autonomous selection often produces inefficient or even erroneous reasoning paths. To make reasoning more reliable and flexible, it is important to develop methods for controlling reasoning strategies. Existing methods struggle to control fine-grained reasoning strategies due to conceptual entanglement in L"},"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":"2601.03595","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-01-07T05:26:26Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"c9eff405471f1e6ad3a975d837e5bdf74a4194a193fca89bcf1d44b541f77a1f","abstract_canon_sha256":"67c0d8df4fd3ab1eccf7f0564b400be4acd80e5425ba55e001e2fbc7c984de32"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-29T00:24:34.848588Z","signature_b64":"Vq/lSDtrLEMz4VGPiR1jHuGAg3cushAGGSYjtfSRMM4wICXYjdYngf2nsp7BU/axY066Dq1rrY13HjuxEg9qBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"08f11ab781443b2a5f3748f7b8fe93459e3d83bacaf31ea41fb83ee960891575","last_reissued_at":"2026-07-29T00:24:34.847515Z","signature_status":"signed_v1","first_computed_at":"2026-07-29T00:24:34.847515Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Controllable LLM Reasoning via Sparse Autoencoder-Based Steering","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Boyi Deng, Dayiheng Liu, Fengli Xu, Fuli Feng, Mingfeng Xue, Wenjie Wang, Yi Fang","submitted_at":"2026-01-07T05:26:26Z","abstract_excerpt":"Large Reasoning Models (LRMs) exhibit human-like cognitive reasoning strategies (\\eg backtracking, cross-verification) during the reasoning process, which improves their performance on complex tasks. Currently, reasoning strategies are autonomously selected by LRMs themselves. However, such autonomous selection often produces inefficient or even erroneous reasoning paths. To make reasoning more reliable and flexible, it is important to develop methods for controlling reasoning strategies. Existing methods struggle to control fine-grained reasoning strategies due to conceptual entanglement in L"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2601.03595","kind":"arxiv","version":2},"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/2601.03595/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":"2601.03595","created_at":"2026-07-29T00:24:34.848006+00:00"},{"alias_kind":"arxiv_version","alias_value":"2601.03595v2","created_at":"2026-07-29T00:24:34.848006+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2601.03595","created_at":"2026-07-29T00:24:34.848006+00:00"},{"alias_kind":"pith_short_12","alias_value":"BDYRVN4BIQ5S","created_at":"2026-07-29T00:24:34.848006+00:00"},{"alias_kind":"pith_short_16","alias_value":"BDYRVN4BIQ5SUXZX","created_at":"2026-07-29T00:24:34.848006+00:00"},{"alias_kind":"pith_short_8","alias_value":"BDYRVN4B","created_at":"2026-07-29T00:24:34.848006+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":6,"sample":[{"citing_arxiv_id":"2606.26620","citing_title":"Discovering Millions of Interpretable Features with Sparse Autoencoders","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2606.06188","citing_title":"The Tell-Tale Norm: $\\ell_2$ Magnitude as a Signal for Reasoning Dynamics in Large Language Models","ref_index":16,"is_internal_anchor":true},{"citing_arxiv_id":"2606.00726","citing_title":"Latent Reward Steering: An Adaptive Inference-Time Framework that Implicitly Promotes Cognitive Behaviors in Reasoning LLMs","ref_index":28,"is_internal_anchor":true},{"citing_arxiv_id":"2605.22902","citing_title":"Transcoders Trace Visual Grounding and Hallucinations in Vision-Language Models","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2605.23040","citing_title":"Steered Generation via Gradient-Based Optimization on Sparse Query Features","ref_index":14,"is_internal_anchor":true},{"citing_arxiv_id":"2605.11887","citing_title":"Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Models","ref_index":38,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BDYRVN4BIQ5SUXZXJD33R7UTIW","json":"https://pith.science/pith/BDYRVN4BIQ5SUXZXJD33R7UTIW.json","graph_json":"https://pith.science/api/pith-number/BDYRVN4BIQ5SUXZXJD33R7UTIW/graph.json","events_json":"https://pith.science/api/pith-number/BDYRVN4BIQ5SUXZXJD33R7UTIW/events.json","paper":"https://pith.science/paper/BDYRVN4B"},"agent_actions":{"view_html":"https://pith.science/pith/BDYRVN4BIQ5SUXZXJD33R7UTIW","download_json":"https://pith.science/pith/BDYRVN4BIQ5SUXZXJD33R7UTIW.json","view_paper":"https://pith.science/paper/BDYRVN4B","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2601.03595&json=true","fetch_graph":"https://pith.science/api/pith-number/BDYRVN4BIQ5SUXZXJD33R7UTIW/graph.json","fetch_events":"https://pith.science/api/pith-number/BDYRVN4BIQ5SUXZXJD33R7UTIW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BDYRVN4BIQ5SUXZXJD33R7UTIW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BDYRVN4BIQ5SUXZXJD33R7UTIW/action/storage_attestation","attest_author":"https://pith.science/pith/BDYRVN4BIQ5SUXZXJD33R7UTIW/action/author_attestation","sign_citation":"https://pith.science/pith/BDYRVN4BIQ5SUXZXJD33R7UTIW/action/citation_signature","submit_replication":"https://pith.science/pith/BDYRVN4BIQ5SUXZXJD33R7UTIW/action/replication_record"}},"created_at":"2026-07-29T00:24:34.848006+00:00","updated_at":"2026-07-29T00:24:34.848006+00:00"}