{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:HSYGU7KMA2FLLY7IW27AG3LMB3","short_pith_number":"pith:HSYGU7KM","canonical_record":{"source":{"id":"2605.27701","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-26T21:20:45Z","cross_cats_sorted":[],"title_canon_sha256":"bf51cb67026d59770f62d9efda880571735e27303b3204cb7aa877dcb479c0a4","abstract_canon_sha256":"ce3b2129f3a62243045f6afd7e3cba44eaff72956ecda5639d39a3bd05447aa6"},"schema_version":"1.0"},"canonical_sha256":"3cb06a7d4c068ab5e3e8b6be036d6c0ecac88784ea6c6868940724411b5dd7a0","source":{"kind":"arxiv","id":"2605.27701","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.27701","created_at":"2026-05-28T01:04:46Z"},{"alias_kind":"arxiv_version","alias_value":"2605.27701v1","created_at":"2026-05-28T01:04:46Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.27701","created_at":"2026-05-28T01:04:46Z"},{"alias_kind":"pith_short_12","alias_value":"HSYGU7KMA2FL","created_at":"2026-05-28T01:04:46Z"},{"alias_kind":"pith_short_16","alias_value":"HSYGU7KMA2FLLY7I","created_at":"2026-05-28T01:04:46Z"},{"alias_kind":"pith_short_8","alias_value":"HSYGU7KM","created_at":"2026-05-28T01:04:46Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:HSYGU7KMA2FLLY7IW27AG3LMB3","target":"record","payload":{"canonical_record":{"source":{"id":"2605.27701","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-26T21:20:45Z","cross_cats_sorted":[],"title_canon_sha256":"bf51cb67026d59770f62d9efda880571735e27303b3204cb7aa877dcb479c0a4","abstract_canon_sha256":"ce3b2129f3a62243045f6afd7e3cba44eaff72956ecda5639d39a3bd05447aa6"},"schema_version":"1.0"},"canonical_sha256":"3cb06a7d4c068ab5e3e8b6be036d6c0ecac88784ea6c6868940724411b5dd7a0","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:04:46.622647Z","signature_b64":"SEpHUEO1S22BZ7vVhmql387FnDBbbhSeKLVzOLpVpnwVG+knxwQXN9qENqaVbZnNxULYCCD7R8NWYbX5gftuDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3cb06a7d4c068ab5e3e8b6be036d6c0ecac88784ea6c6868940724411b5dd7a0","last_reissued_at":"2026-05-28T01:04:46.622229Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:04:46.622229Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.27701","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":"2026-05-28T01:04:46Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"r8WXDU1wkCdWp5FJalPV9veZUoJQiyWBjgIhZitcNuWoUwnOA0A3z1Bf6EFVa7GtOqNqpv+L0fuL80JLYBH1Cw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-30T03:55:08.072848Z"},"content_sha256":"656c7f76bc5469ccfa08f8fb677309dc8f9efa92b2ff72e14450ea885a6a8da4","schema_version":"1.0","event_id":"sha256:656c7f76bc5469ccfa08f8fb677309dc8f9efa92b2ff72e14450ea885a6a8da4"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:HSYGU7KMA2FLLY7IW27AG3LMB3","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Cross-Entropy Games and Frost Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Arthur Renard, Cl\\'ement Hongler, Franck Gabriel, Valentin Hartmann","submitted_at":"2026-05-26T21:20:45Z","abstract_excerpt":"We present Frost Training, a method for improving Monte Carlo-based policy optimization for a large family of LLM-as-a-judge tasks called Cross-Entropy Games. The key idea is to exploit the gradient of the reward function in embedding space. This signal is used in the Greedy Coordinate Gradient (GCG) jailbreaking technique; we demonstrate for the first time that it can also be used to boost model training. We validate our method using GRPO training for maximum-likelihood infilling. Frost Training improves the model's ability to generate high-scoring outputs, reaching higher maximum scores in a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.27701","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/2605.27701/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":"2026-05-28T01:04:46Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"3ZbR4rVZWER2NWgM7PHwKdmz1tbEsI4a3AtCfVNlpkEqtQw80B9JcrYL/fhlSILjbsZ4KNDYLp5fd9qk3UV0CA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-30T03:55:08.073233Z"},"content_sha256":"e576bf498e5242490d24675610b3c64cb4efc1793576d84e7a6344fc1bbaff37","schema_version":"1.0","event_id":"sha256:e576bf498e5242490d24675610b3c64cb4efc1793576d84e7a6344fc1bbaff37"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/HSYGU7KMA2FLLY7IW27AG3LMB3/bundle.json","state_url":"https://pith.science/pith/HSYGU7KMA2FLLY7IW27AG3LMB3/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/HSYGU7KMA2FLLY7IW27AG3LMB3/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":"2026-05-30T03:55:08Z","links":{"resolver":"https://pith.science/pith/HSYGU7KMA2FLLY7IW27AG3LMB3","bundle":"https://pith.science/pith/HSYGU7KMA2FLLY7IW27AG3LMB3/bundle.json","state":"https://pith.science/pith/HSYGU7KMA2FLLY7IW27AG3LMB3/state.json","well_known_bundle":"https://pith.science/.well-known/pith/HSYGU7KMA2FLLY7IW27AG3LMB3/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:HSYGU7KMA2FLLY7IW27AG3LMB3","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":"ce3b2129f3a62243045f6afd7e3cba44eaff72956ecda5639d39a3bd05447aa6","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-26T21:20:45Z","title_canon_sha256":"bf51cb67026d59770f62d9efda880571735e27303b3204cb7aa877dcb479c0a4"},"schema_version":"1.0","source":{"id":"2605.27701","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.27701","created_at":"2026-05-28T01:04:46Z"},{"alias_kind":"arxiv_version","alias_value":"2605.27701v1","created_at":"2026-05-28T01:04:46Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.27701","created_at":"2026-05-28T01:04:46Z"},{"alias_kind":"pith_short_12","alias_value":"HSYGU7KMA2FL","created_at":"2026-05-28T01:04:46Z"},{"alias_kind":"pith_short_16","alias_value":"HSYGU7KMA2FLLY7I","created_at":"2026-05-28T01:04:46Z"},{"alias_kind":"pith_short_8","alias_value":"HSYGU7KM","created_at":"2026-05-28T01:04:46Z"}],"graph_snapshots":[{"event_id":"sha256:e576bf498e5242490d24675610b3c64cb4efc1793576d84e7a6344fc1bbaff37","target":"graph","created_at":"2026-05-28T01:04:46Z","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/2605.27701/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We present Frost Training, a method for improving Monte Carlo-based policy optimization for a large family of LLM-as-a-judge tasks called Cross-Entropy Games. The key idea is to exploit the gradient of the reward function in embedding space. This signal is used in the Greedy Coordinate Gradient (GCG) jailbreaking technique; we demonstrate for the first time that it can also be used to boost model training. We validate our method using GRPO training for maximum-likelihood infilling. Frost Training improves the model's ability to generate high-scoring outputs, reaching higher maximum scores in a","authors_text":"Arthur Renard, Cl\\'ement Hongler, Franck Gabriel, Valentin Hartmann","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-26T21:20:45Z","title":"Cross-Entropy Games and Frost Training"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.27701","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":"sha256:656c7f76bc5469ccfa08f8fb677309dc8f9efa92b2ff72e14450ea885a6a8da4","target":"record","created_at":"2026-05-28T01:04:46Z","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":"ce3b2129f3a62243045f6afd7e3cba44eaff72956ecda5639d39a3bd05447aa6","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-26T21:20:45Z","title_canon_sha256":"bf51cb67026d59770f62d9efda880571735e27303b3204cb7aa877dcb479c0a4"},"schema_version":"1.0","source":{"id":"2605.27701","kind":"arxiv","version":1}},"canonical_sha256":"3cb06a7d4c068ab5e3e8b6be036d6c0ecac88784ea6c6868940724411b5dd7a0","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"3cb06a7d4c068ab5e3e8b6be036d6c0ecac88784ea6c6868940724411b5dd7a0","first_computed_at":"2026-05-28T01:04:46.622229Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-28T01:04:46.622229Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"SEpHUEO1S22BZ7vVhmql387FnDBbbhSeKLVzOLpVpnwVG+knxwQXN9qENqaVbZnNxULYCCD7R8NWYbX5gftuDg==","signature_status":"signed_v1","signed_at":"2026-05-28T01:04:46.622647Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.27701","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:656c7f76bc5469ccfa08f8fb677309dc8f9efa92b2ff72e14450ea885a6a8da4","sha256:e576bf498e5242490d24675610b3c64cb4efc1793576d84e7a6344fc1bbaff37"],"state_sha256":"5697bfd47a49f544ae48a6382cafd3acf476b94fb11e9af59a04b8db6481870e"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"kUERqCt68j6zn/IiZsxv4vn6HDWTH4VbIxrHpZrZkQPZAyBIfNGUj5APcjcGR14XXQ5gEkmLatVELERQuXsLCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-30T03:55:08.076773Z","bundle_sha256":"4373e96ce2a04112d4845a1371eb355ad6b9e6100c567b70a37a110a19626306"}}