{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:FGDSKN473HL4NEH7AXDK4NEUEX","short_pith_number":"pith:FGDSKN47","canonical_record":{"source":{"id":"2508.16546","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.LG","submitted_at":"2025-08-22T17:10:37Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"297c7addee0d1385b07e4805b92aac0b21a0d812bed4193a26c64ed4fb850666","abstract_canon_sha256":"148befcaa82b8769f0259cf1cbb242e3c70548ed9792aff45a0ca7be782771a6"},"schema_version":"1.0"},"canonical_sha256":"298725379fd9d7c690ff05c6ae349425de5de04c0ba5aee7e0b3579ded2d0b5c","source":{"kind":"arxiv","id":"2508.16546","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2508.16546","created_at":"2026-07-05T11:57:50Z"},{"alias_kind":"arxiv_version","alias_value":"2508.16546v1","created_at":"2026-07-05T11:57:50Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.16546","created_at":"2026-07-05T11:57:50Z"},{"alias_kind":"pith_short_12","alias_value":"FGDSKN473HL4","created_at":"2026-07-05T11:57:50Z"},{"alias_kind":"pith_short_16","alias_value":"FGDSKN473HL4NEH7","created_at":"2026-07-05T11:57:50Z"},{"alias_kind":"pith_short_8","alias_value":"FGDSKN47","created_at":"2026-07-05T11:57:50Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:FGDSKN473HL4NEH7AXDK4NEUEX","target":"record","payload":{"canonical_record":{"source":{"id":"2508.16546","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.LG","submitted_at":"2025-08-22T17:10:37Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"297c7addee0d1385b07e4805b92aac0b21a0d812bed4193a26c64ed4fb850666","abstract_canon_sha256":"148befcaa82b8769f0259cf1cbb242e3c70548ed9792aff45a0ca7be782771a6"},"schema_version":"1.0"},"canonical_sha256":"298725379fd9d7c690ff05c6ae349425de5de04c0ba5aee7e0b3579ded2d0b5c","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:57:50.531154Z","signature_b64":"bQ7b4gSj4OWb8vpr0czhlweR58lHmq1whuqXrfo8fH/O0cX5Im4a/6NAjI0C58W7TXx3KZxEuTLJk+bfN3aCDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"298725379fd9d7c690ff05c6ae349425de5de04c0ba5aee7e0b3579ded2d0b5c","last_reissued_at":"2026-07-05T11:57:50.530665Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:57:50.530665Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2508.16546","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:57:50Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"+w8OZyq/nbSA2yoXE5l9CUALtTWjLZ8kSJ92QRzg1uo2HXWhC58A8gk/VQxOquXONqeLE0cX5QNWITO02yOQAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T01:16:21.959761Z"},"content_sha256":"d9a8bab9738c8fd843d9d48b08d809131b28d121fcb9472e7805e972dea90158","schema_version":"1.0","event_id":"sha256:d9a8bab9738c8fd843d9d48b08d809131b28d121fcb9472e7805e972dea90158"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:FGDSKN473HL4NEH7AXDK4NEUEX","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"RL Is Neither a Panacea Nor a Mirage: Understanding Supervised vs. Reinforcement Learning Fine-Tuning for LLMs","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Hangzhan Jin, Mohammad Hamdaqa, Sicheng Lv, Sifan Wu","submitted_at":"2025-08-22T17:10:37Z","abstract_excerpt":"Training large language models (LLMs) from scratch is increasingly impractical, making post-training methods such as supervised fine-tuning (SFT) and reinforcement-learning fine-tuning (RL-FT, e.g., PPO) central to modern practice. Using an out-of-distribution (OOD) variant of the 24-point card game and new spectrum-based diagnostics, we revisit how these two stages reshape model representation and OOD performance. Our key findings are- (1) RL-FT can restore much of the OOD performance loss from SFT (e.g., Llama-11B 8.97% to 15.38%, Qwen-7B 17.09% to 19.66%). But when SFT induces severe overfi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.16546","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/2508.16546/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:57:50Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"zr2ROH6JA6V+PUOjc2KBNy8V5fu+fevZcl3LLVdOe1oEPSW1zkOV3BZ/HUe6kTyIuu2jxPnJkgsWYFfuJ3HyCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T01:16:21.960259Z"},"content_sha256":"7f34749346068f90ceb41de60b0c32190c76055c9957d31a94a19f01987a13a6","schema_version":"1.0","event_id":"sha256:7f34749346068f90ceb41de60b0c32190c76055c9957d31a94a19f01987a13a6"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/FGDSKN473HL4NEH7AXDK4NEUEX/bundle.json","state_url":"https://pith.science/pith/FGDSKN473HL4NEH7AXDK4NEUEX/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/FGDSKN473HL4NEH7AXDK4NEUEX/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-08T01:16:21Z","links":{"resolver":"https://pith.science/pith/FGDSKN473HL4NEH7AXDK4NEUEX","bundle":"https://pith.science/pith/FGDSKN473HL4NEH7AXDK4NEUEX/bundle.json","state":"https://pith.science/pith/FGDSKN473HL4NEH7AXDK4NEUEX/state.json","well_known_bundle":"https://pith.science/.well-known/pith/FGDSKN473HL4NEH7AXDK4NEUEX/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:FGDSKN473HL4NEH7AXDK4NEUEX","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":"148befcaa82b8769f0259cf1cbb242e3c70548ed9792aff45a0ca7be782771a6","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.LG","submitted_at":"2025-08-22T17:10:37Z","title_canon_sha256":"297c7addee0d1385b07e4805b92aac0b21a0d812bed4193a26c64ed4fb850666"},"schema_version":"1.0","source":{"id":"2508.16546","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2508.16546","created_at":"2026-07-05T11:57:50Z"},{"alias_kind":"arxiv_version","alias_value":"2508.16546v1","created_at":"2026-07-05T11:57:50Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.16546","created_at":"2026-07-05T11:57:50Z"},{"alias_kind":"pith_short_12","alias_value":"FGDSKN473HL4","created_at":"2026-07-05T11:57:50Z"},{"alias_kind":"pith_short_16","alias_value":"FGDSKN473HL4NEH7","created_at":"2026-07-05T11:57:50Z"},{"alias_kind":"pith_short_8","alias_value":"FGDSKN47","created_at":"2026-07-05T11:57:50Z"}],"graph_snapshots":[{"event_id":"sha256:7f34749346068f90ceb41de60b0c32190c76055c9957d31a94a19f01987a13a6","target":"graph","created_at":"2026-07-05T11:57:50Z","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":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2508.16546/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Training large language models (LLMs) from scratch is increasingly impractical, making post-training methods such as supervised fine-tuning (SFT) and reinforcement-learning fine-tuning (RL-FT, e.g., PPO) central to modern practice. Using an out-of-distribution (OOD) variant of the 24-point card game and new spectrum-based diagnostics, we revisit how these two stages reshape model representation and OOD performance. Our key findings are- (1) RL-FT can restore much of the OOD performance loss from SFT (e.g., Llama-11B 8.97% to 15.38%, Qwen-7B 17.09% to 19.66%). But when SFT induces severe overfi","authors_text":"Hangzhan Jin, Mohammad Hamdaqa, Sicheng Lv, Sifan Wu","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.LG","submitted_at":"2025-08-22T17:10:37Z","title":"RL Is Neither a Panacea Nor a Mirage: Understanding Supervised vs. Reinforcement Learning Fine-Tuning for LLMs"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.16546","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:d9a8bab9738c8fd843d9d48b08d809131b28d121fcb9472e7805e972dea90158","target":"record","created_at":"2026-07-05T11:57:50Z","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":"148befcaa82b8769f0259cf1cbb242e3c70548ed9792aff45a0ca7be782771a6","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.LG","submitted_at":"2025-08-22T17:10:37Z","title_canon_sha256":"297c7addee0d1385b07e4805b92aac0b21a0d812bed4193a26c64ed4fb850666"},"schema_version":"1.0","source":{"id":"2508.16546","kind":"arxiv","version":1}},"canonical_sha256":"298725379fd9d7c690ff05c6ae349425de5de04c0ba5aee7e0b3579ded2d0b5c","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"298725379fd9d7c690ff05c6ae349425de5de04c0ba5aee7e0b3579ded2d0b5c","first_computed_at":"2026-07-05T11:57:50.530665Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:57:50.530665Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"bQ7b4gSj4OWb8vpr0czhlweR58lHmq1whuqXrfo8fH/O0cX5Im4a/6NAjI0C58W7TXx3KZxEuTLJk+bfN3aCDQ==","signature_status":"signed_v1","signed_at":"2026-07-05T11:57:50.531154Z","signed_message":"canonical_sha256_bytes"},"source_id":"2508.16546","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:d9a8bab9738c8fd843d9d48b08d809131b28d121fcb9472e7805e972dea90158","sha256:7f34749346068f90ceb41de60b0c32190c76055c9957d31a94a19f01987a13a6"],"state_sha256":"ef0fd92d0a6cd465827875eea856f68c4a910aa55957b252126659aff7434574"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6Vh9rfs7z0MD9Y21X43fA9SFud/Rx0aw+4vg5j1ojjGDQKiEaTe/Ljt8/hqWfSkhc/l/Bxb8pwY0YpbFoTD/Ag==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-08T01:16:21.966437Z","bundle_sha256":"a8fb86841d35ee71ef239f17c7e007f4cd54889f8033d345d91b20dbb34e74dc"}}