{"_canonicalization":{"envelope_id":"axm_ + sha256(envelope minus {signature, axiom_id, anchors})","envelope_signature":"ed25519(envelope minus {signature, axiom_id})","json":"sort_keys=True, separators=(',',':'), ensure_ascii=False, allow_nan=False, utf-8","leaf_hash":"sha256(0x00 || canonical_json(envelope_full))","seal_signature":"ed25519(seal minus {signature, sig_algorithm})"},"axiom_id":"axm_30adc4c6f45a67781b71fb518eda3c051fa4e197a66676eafa1ba65cd6735f0a","bitcoin_anchor":{"bitcoin_attestations":[],"calendar_attestations":[],"ots_url":"","stamped_at":"","status":"pending_next_stamp"},"envelope":{"anchors":[{"chain":"crovia.axiom_graph","height":0,"merkle_proof":"spider_vendor_press_v1","root_at_anchor":"spider_vendor_press_v1"}],"axiom_id":"axm_30adc4c6f45a67781b71fb518eda3c051fa4e197a66676eafa1ba65cd6735f0a","axiom_type":"AX.OBS","body":{"axiom_subtype":"news.vendor_press.v1","category":"news","fingerprint":"6f41badf219795dfbb40670ce81c7b5b2e99f75ece15bd8621c05a4add6cb216","published":"Wed, 20 May 2026 00:00:00 -0400","receipt_hash":"6f41badf219795dfbb40670ce81c7b5b2e99f75ece15bd8621c05a4add6cb216","schema":"spider.news.vendor_press.v1","spider":"vendor_press","spider_record":{"axiom_subtype":"news.vendor_press.v1","category":"news","decision_hint":"POSITIVE","envelope_target":"AX.OBS","fingerprint":"6f41badf219795dfbb40670ce81c7b5b2e99f75ece15bd8621c05a4add6cb216","observed_at":"2026-05-20T04:43:44.562035Z","parent_run_hash":"5f904c2c2fecb6b44f2adce8bdc9de914b9a39f7c4fdd7bf086a6c2361a30f8c","published":"Wed, 20 May 2026 00:00:00 -0400","runtime_version":"0.1.0","schema":"spider.news.vendor_press.v1","source_status":200,"source_url":"https://export.arxiv.org/rss/cs.AI","spider":"vendor_press","summary_excerpt":"arXiv:2512.23461v2 Announce Type: replace-cross \nAbstract: Reward models (RMs) are essential in reinforcement learning from human feedback (RLHF) to align large language models (LLMs) with human values. However, RM training data is commonly recognized as low-quality, containing inductive biases that can easily lead to overfitting and reward hacking. For example, more detailed and comprehensive responses are usually human-preferred but with more words, leading response length to become one of the inevitable inductive biases. A limited number of prior RM debiasing approaches either target a single specific type of bias or model the problem with only simple linear correlations, \\textit{e.g.}, Pearson coefficients. To mitigate more complex and diverse inductive biases in reward modeling, we introduce a novel information-theoretic debiasing method called \\textbf{D}ebiasing via \\textbf{I}nformation optimization for \\textbf{R}M (DIR). Inspired by the information bottleneck (IB), we maximize t","title":"Eliminating Inductive Bias in Reward Models with Information-Theoretic Guidance","url":"https://arxiv.org/abs/2512.23461","vendor":"arxiv_cs_ai"},"summary":"arXiv:2512.23461v2 Announce Type: replace-cross \nAbstract: Reward models (RMs) are essential in reinforcement learning from human feedback (RLHF) to align large language models (LLMs) with human values. However, RM training data is commonly recognized as low-quality, containing inductive biases that can easily lead to overfitting and reward hacking. For example, more detailed and comprehensive responses are usually human-preferred but with more words, leading response length to become one of the inevitable inductive biases. A limited number of prior RM debiasing approaches either target a single specific type of bias or model the problem with only simple linear correlations, \\textit{e.g.}, Pearson coefficients. To mitigate more complex and diverse inductive biases in reward modeling, we introduce a novel information-theoretic debiasing method called \\textbf{D}ebiasing via \\textbf{I}nformation optimization for \\textbf{R}M (DIR). Inspired by the information bottleneck (IB), we maximize t","title":"Eliminating Inductive Bias in Reward Models with Information-Theoretic Guidance","vendor":"arxiv_cs_ai"},"confidence":{"method":"deterministic"},"decision":"POSITIVE","issued_at":"2026-05-20T04:43:44Z","notes":"Spider vendor_press (news) news.vendor_press.v1","object":{"captured_by":"crovia.spider.vendor_press","primary_source_url":"https://arxiv.org/abs/2512.23461"},"predecessors":[],"schema":"crovia.axiom.v1","signature":"ed25519:e6d85182baf998ce29dd651945434b63edb64beccf5eca98bb988f5fece47773eb410352dcbc83f704b55e1f22f3377cf68bdcd4a4c1f833e8b1bb3168fcac0f","signer":"crovia.substrate","subject":{"observed_at":"2026-05-20T04:43:44Z","source_collector":"spider:vendor_press","target_id":"https://arxiv.org/abs/2512.23461"},"tsa":{"authority":"crovia.substrate.bootstrap","rfc3161_token":"{\"kind\":\"crovia.bootstrap.tsa\",\"source_jsonl\":\"/opt/crovia/spider/data/news/vendor_press_v1.jsonl\",\"source_seal_merkle_root\":\"spider_vendor_press_v1\",\"upgrade_path\":\"Sessione H \\u2014 OpenTimestamps weekly anchor\"}"},"zk_mode":"clear","zk_proof":null},"ledger":{"leaf_hash":"059ec7e54ee4c47beb5793d06fd2d5113bd8acd1b4a50d21753951bb544d1dc0","leaf_index":145277,"ledger_path":"/opt/crovia/substrate/axiom_ledger.jsonl"},"merkle_proof":{"hash_alg":"sha256","leaf_prefix":"0x00","node_prefix":"0x01","odd_leaf_rule":"duplicate_last","path":[{"sibling":"4fea8b0ab8d1456be88c3259d903c01006af6d53e08adabac35fcd8bb5afd565","side":"left"},{"sibling":"e600cb8f3af07f98758da5af8ae163729ba857953ea36c99190d095a67bcda1e","side":"right"},{"sibling":"cf39e26f12c0fab82ccff272a963390fcf7d10640b03af455fe8efd8e9f165df","side":"left"},{"sibling":"e8c60014467dbadd2e2d75d812cae715f9bbb0a221af6ea5c724cec29d57156a","side":"left"},{"sibling":"2d39257204060484eb4366873a43f6b2fe4aaacf5433657f39607d37de515cd7","side":"left"},{"sibling":"b891f8102dbec11b5872d47b2e640beca78f5f15b3e6d01c8d7bbebe1e3b209f","side":"left"},{"sibling":"8abfb0c0cfeb30fba16eb01548a8b7ce35f077e44d727cc0fe4a62d471800c6e","side":"left"},{"sibling":"79fb6d27e8a49dd15a97fca96f52d62ff5ee7213d44a5526459d33f42a050928","side":"right"},{"sibling":"f073aa7be27ee7e1d9eb0de5f129f7e9bae0c584fb959378c395b65831dc1d1d","side":"left"},{"sibling":"1526885f19d1fadf6955cf519dbc4e62d593a4bba99d741e8c301740a7068233","side":"left"},{"sibling":"e3a7d5c07f161682d61bd453ffc02ecdf87cfeda70f986d6650017f9d2d6b265","side":"left"},{"sibling":"edbc49f08e5b92291934c05c9e6efd270a6b0698d8d2fa474006366027dfe098","side":"right"},{"sibling":"3e4df6e7457cecbf36f350375e72dcab336a3984422e4c406ef809e4e2944e96","side":"left"},{"sibling":"8f4c0fbe56b6c010fbb8c782ebcd478079bb3f991d8704e2534209a075d9163c","side":"left"},{"sibling":"be08fedc4e72a6fb56606f66812fae7317e09690b9acb18385f4ab117a981238","side":"right"},{"sibling":"0534329a7475dc9df51998c83c16892126679dade0fa34182f21e869599386c7","side":"right"},{"sibling":"2d24720928ead0e7670650eb55f558c4f20e4c18df376f47ba72cfa8cf0ed344","side":"right"},{"sibling":"d841ad93efda0869e5eb97678f348f03f5caab4353e05ff4bf18f47fb945b822","side":"left"}]},"schema":"crovia.axiom_proof.v1","seal":{"first_collector_run_id":"","first_receipt_hash":"","jsonl_path":"/opt/crovia/substrate/axiom_ledger.jsonl","key_id":"430895f101d38164","last_collector_run_id":"","last_receipt_hash":"","leaf_count":147301,"merkle_root":"08903d7159c3b38eeeeafc09eab15139ea417f1d94f02f1fbc87296b37db840a","public_key_hex":"cf742e26f75669dc673cb5c0786a1ae23ae8ca19c347317192ce40c28a7ff25c","run_id":"hourly_json_retrofit_20260521T183701Z","schema":"crovia.seal.v1","seal_family_version":"crovia-seal-family/1","seal_kind":"substrate_batch","sealed_at":"2026-05-21T18:37:33Z","sig_algorithm":"ed25519","signature":"905f2924632dfa2970c8690285f5b5d4a1d891d0e0ef1cbc404ebec2fd937215ac768e16f0a9f28b18977a55ae0bfd226db6833ae7ef588729054117d2da7303","signer_version":"1.1.0"},"trust_root":{"key_id":"430895f101d38164","public_key_hex":"cf742e26f75669dc673cb5c0786a1ae23ae8ca19c347317192ce40c28a7ff25c","signature_algorithm":"ed25519","url":"/registry/canon/TRUST_ROOT.md"},"verifier":{"spec":"/registry/canon/AXIOM_RECEIPT_v1.md","url":"/v/axm_30adc4c6f45a67781b71fb518eda3c051fa4e197a66676eafa1ba65cd6735f0a"}}