{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:63NRELHFFN5IYRD4CHKJSYLPIL","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":"0d7b2fceed870fc9915bb543fe1bc4aad71a2254498f74720408394b6965a165","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.GT","submitted_at":"2026-05-08T06:56:35Z","title_canon_sha256":"ec01d02e8ddaea2090a0c9969382906a7bd8f682e5499ce7de30b309fee65ba3"},"schema_version":"1.0","source":{"id":"2605.13875","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.13875","created_at":"2026-05-17T23:39:19Z"},{"alias_kind":"arxiv_version","alias_value":"2605.13875v1","created_at":"2026-05-17T23:39:19Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.13875","created_at":"2026-05-17T23:39:19Z"},{"alias_kind":"pith_short_12","alias_value":"63NRELHFFN5I","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"63NRELHFFN5IYRD4","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"63NRELHF","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:ed136807ceb5e28b334e297753e4aa49cb1d33de084418e50e4a5291704965e8","target":"graph","created_at":"2026-05-17T23:39:19Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"CAGE enables flexible and fine-grained trade-offs across objectives at inference time, consistently outperforming existing test-time alignment methods while requiring no retraining. It further supports weak-to-strong generalization, making multi-objective alignment practical in resource-constrained settings."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That modeling heterogeneous objectives as strategic principals allocating token-level incentives produces an equilibrium policy whose joint effect meaningfully captures real user preferences, and that the EPEC-based algorithm reliably computes this equilibrium with the claimed existence, uniqueness, convergence, and stability guarantees."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"CAGE uses common-agency games and an EPEC algorithm to compute equilibrium policies that balance multiple conflicting objectives for test-time LLM alignment."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"CAGE treats multiple conflicting alignment goals as strategic principals bidding token incentives to produce an equilibrium LLM policy at inference time."}],"snapshot_sha256":"ba3aeb72703cb6e24891c5dae8672767889a68094fa699046c184755d7700103"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"aebd8280c17417da790c7bdeedd992a81e4524c1d32020d083c0fcbfb65b478e"},"paper":{"abstract_excerpt":"Aligning large language models (LLMs) with human preferences is inherently multi-objective: different users and evaluation criteria impose heterogeneous and often conflicting requirements on model outputs. We propose CAGE (Common-Agency Games for Alignment), a training-free, game-theoretic framework for multi-objective test-time alignment. CAGE models alignment objectives as strategic principals that allocate token-level incentives to a shared LLM, inducing an equilibrium policy that captures the joint effect of competing objectives. We develop an efficient algorithm based on equilibrium probl","authors_text":"Baiting Chen, Rui Yu, Tong Zhu, Xiaowu Dai","cross_cats":[],"headline":"CAGE treats multiple conflicting alignment goals as strategic principals bidding token incentives to produce an equilibrium LLM policy at inference time.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.GT","submitted_at":"2026-05-08T06:56:35Z","title":"Common-agency Games for Multi-Objective Test-Time Alignment"},"references":{"count":230,"internal_anchors":19,"resolved_work":230,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"TinyLlama: An Open-Source Small Language Model , author=. 2024 , eprint=","work_id":"4b2a85e5-7934-4e57-ba32-e7ac0a26d388","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"International Conference on Learning Representations , year=","work_id":"7d66a0d4-a1e1-4771-b0cd-ea8c424cb4bd","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"The Twelfth International Conference on Learning Representations , year=","work_id":"44d6bf99-5902-43da-b8fd-83297976c6e0","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"C hat D ev: Communicative agents for software development","work_id":"166b34d2-9477-4b53-b493-bcce91575006","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Encouraging divergent thinking in large language models through multi-agent debate","work_id":"5e657239-8ab5-4c3c-ac58-63b4c9d2e420","year":2024}],"snapshot_sha256":"0d6a0ec0b4c4362b2aaaab7e88e2dcbc27332505742a1851b0d20041ea2a4151"},"source":{"id":"2605.13875","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-15T06:11:14.544441Z","id":"1c5a5ac2-e479-4d34-af9e-4253368271a2","model_set":{"reader":"grok-4.3"},"one_line_summary":"CAGE uses common-agency games and an EPEC algorithm to compute equilibrium policies that balance multiple conflicting objectives for test-time LLM alignment.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"CAGE treats multiple conflicting alignment goals as strategic principals bidding token incentives to produce an equilibrium LLM policy at inference time.","strongest_claim":"CAGE enables flexible and fine-grained trade-offs across objectives at inference time, consistently outperforming existing test-time alignment methods while requiring no retraining. It further supports weak-to-strong generalization, making multi-objective alignment practical in resource-constrained settings.","weakest_assumption":"That modeling heterogeneous objectives as strategic principals allocating token-level incentives produces an equilibrium policy whose joint effect meaningfully captures real user preferences, and that the EPEC-based algorithm reliably computes this equilibrium with the claimed existence, uniqueness, convergence, and stability guarantees."}},"verdict_id":"1c5a5ac2-e479-4d34-af9e-4253368271a2"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:b7cf9e4885c957d4c2516545a097d1a414b8470e9dc293ad626183c79024ec7b","target":"record","created_at":"2026-05-17T23:39:19Z","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":"0d7b2fceed870fc9915bb543fe1bc4aad71a2254498f74720408394b6965a165","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.GT","submitted_at":"2026-05-08T06:56:35Z","title_canon_sha256":"ec01d02e8ddaea2090a0c9969382906a7bd8f682e5499ce7de30b309fee65ba3"},"schema_version":"1.0","source":{"id":"2605.13875","kind":"arxiv","version":1}},"canonical_sha256":"f6db122ce52b7a8c447c11d499616f42fbb5cdaaa2db7b6b45e455121da19756","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"f6db122ce52b7a8c447c11d499616f42fbb5cdaaa2db7b6b45e455121da19756","first_computed_at":"2026-05-17T23:39:19.262842Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:39:19.262842Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"WzcrMuPSQSdee8hb6nNyW8rcTfw7z55U7gkmVVTJKWiizLKZ2B75LUGODvyhrGPG5dlzFtE7vWJh3HPZoiQvBg==","signature_status":"signed_v1","signed_at":"2026-05-17T23:39:19.263731Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.13875","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:b7cf9e4885c957d4c2516545a097d1a414b8470e9dc293ad626183c79024ec7b","sha256:ed136807ceb5e28b334e297753e4aa49cb1d33de084418e50e4a5291704965e8"],"state_sha256":"2ae21f27ea0cc6c27d6ada9a03606585776fc051c079e4c4fcdd37ee1fe4dc1b"}