{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:S3XJGU7YUBVUTFUF4E3ALTSRU5","short_pith_number":"pith:S3XJGU7Y","schema_version":"1.0","canonical_sha256":"96ee9353f8a06b499685e13605ce51a775be123a87d67298a24495492b0e9af8","source":{"kind":"arxiv","id":"2606.00674","version":1},"attestation_state":"computed","paper":{"title":"The Paradox of Outcome Optimization: A Causal Information-Theoretic Bound on Reasoning Shortcuts in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Wenxiang Geng, Yiming Zhang, Yining Sun, Zenghui Ding, Zihan Chen","submitted_at":"2026-05-30T11:06:13Z","abstract_excerpt":"Large Language Models (LLMs) aligned via outcome-based Reinforcement Learning (RL) frequently exhibit a critical failure mode: they achieve high performance on in-distribution benchmarks while demonstrating brittle reasoning capabilities on out-of-distribution (OOD) tasks. We term this phenomenon Reward-Induced Manifold Collapse. We establish a theoretical framework bridging Structural Causal Models (SCM) and the Information Bottleneck (IB) principle to explain this paradox. We define reasoning as a high-complexity causal process and shortcut learning as the exploitation of low-complexity spur"},"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":"2606.00674","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-30T11:06:13Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"2f29d9005fe092838e7bee8f12a09017099b86ad1a5cad6954ab8bf30d246e12","abstract_canon_sha256":"bbd28f19ec89445844bb186b6dd2d740f7328298e430837b79aefb36f3d69759"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T01:04:02.367495Z","signature_b64":"yLWphFYBSC64gL77Y7w7f/YtryX847zMetiv62WeewV5bERdbQpjoIqoCr9iS2JSsw/HQFBbl/MzeqrUa6H/Dw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"96ee9353f8a06b499685e13605ce51a775be123a87d67298a24495492b0e9af8","last_reissued_at":"2026-06-02T01:04:02.367042Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T01:04:02.367042Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Paradox of Outcome Optimization: A Causal Information-Theoretic Bound on Reasoning Shortcuts in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Wenxiang Geng, Yiming Zhang, Yining Sun, Zenghui Ding, Zihan Chen","submitted_at":"2026-05-30T11:06:13Z","abstract_excerpt":"Large Language Models (LLMs) aligned via outcome-based Reinforcement Learning (RL) frequently exhibit a critical failure mode: they achieve high performance on in-distribution benchmarks while demonstrating brittle reasoning capabilities on out-of-distribution (OOD) tasks. We term this phenomenon Reward-Induced Manifold Collapse. We establish a theoretical framework bridging Structural Causal Models (SCM) and the Information Bottleneck (IB) principle to explain this paradox. We define reasoning as a high-complexity causal process and shortcut learning as the exploitation of low-complexity spur"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.00674","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/2606.00674/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":"2606.00674","created_at":"2026-06-02T01:04:02.367104+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.00674v1","created_at":"2026-06-02T01:04:02.367104+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.00674","created_at":"2026-06-02T01:04:02.367104+00:00"},{"alias_kind":"pith_short_12","alias_value":"S3XJGU7YUBVU","created_at":"2026-06-02T01:04:02.367104+00:00"},{"alias_kind":"pith_short_16","alias_value":"S3XJGU7YUBVUTFUF","created_at":"2026-06-02T01:04:02.367104+00:00"},{"alias_kind":"pith_short_8","alias_value":"S3XJGU7Y","created_at":"2026-06-02T01:04:02.367104+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/S3XJGU7YUBVUTFUF4E3ALTSRU5","json":"https://pith.science/pith/S3XJGU7YUBVUTFUF4E3ALTSRU5.json","graph_json":"https://pith.science/api/pith-number/S3XJGU7YUBVUTFUF4E3ALTSRU5/graph.json","events_json":"https://pith.science/api/pith-number/S3XJGU7YUBVUTFUF4E3ALTSRU5/events.json","paper":"https://pith.science/paper/S3XJGU7Y"},"agent_actions":{"view_html":"https://pith.science/pith/S3XJGU7YUBVUTFUF4E3ALTSRU5","download_json":"https://pith.science/pith/S3XJGU7YUBVUTFUF4E3ALTSRU5.json","view_paper":"https://pith.science/paper/S3XJGU7Y","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.00674&json=true","fetch_graph":"https://pith.science/api/pith-number/S3XJGU7YUBVUTFUF4E3ALTSRU5/graph.json","fetch_events":"https://pith.science/api/pith-number/S3XJGU7YUBVUTFUF4E3ALTSRU5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/S3XJGU7YUBVUTFUF4E3ALTSRU5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/S3XJGU7YUBVUTFUF4E3ALTSRU5/action/storage_attestation","attest_author":"https://pith.science/pith/S3XJGU7YUBVUTFUF4E3ALTSRU5/action/author_attestation","sign_citation":"https://pith.science/pith/S3XJGU7YUBVUTFUF4E3ALTSRU5/action/citation_signature","submit_replication":"https://pith.science/pith/S3XJGU7YUBVUTFUF4E3ALTSRU5/action/replication_record"}},"created_at":"2026-06-02T01:04:02.367104+00:00","updated_at":"2026-06-02T01:04:02.367104+00:00"}