{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:UO6GJSGJTZ5JBRAHLNP2QN73KM","short_pith_number":"pith:UO6GJSGJ","schema_version":"1.0","canonical_sha256":"a3bc64c8c99e7a90c4075b5fa837fb533afdd70d7198e3cf847aaafb04605626","source":{"kind":"arxiv","id":"2410.21052","version":3},"attestation_state":"computed","paper":{"title":"Getting By Goal Misgeneralization With a Little Help From a Mentor","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Benjamin Plaut, Mohamad H. Danesh, Nguyen X. Khanh, Tu Trinh","submitted_at":"2024-10-28T14:07:41Z","abstract_excerpt":"While reinforcement learning (RL) agents often perform well during training, they can struggle with distribution shift in real-world deployments. One particularly severe risk of distribution shift is goal misgeneralization, where the agent learns a proxy goal that coincides with the true goal during training but not during deployment. In this paper, we explore whether allowing an agent to ask for help from a supervisor in unfamiliar situations can mitigate this issue. We focus on agents trained with PPO in the CoinRun environment, a setting known to exhibit goal misgeneralization. We evaluate "},"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":"2410.21052","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-28T14:07:41Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"112b2f50edd051fb8583f7c6306edbbb5b22087ce8c020fbd1d012cdc5b758eb","abstract_canon_sha256":"b904a7475888887290dcd05942d3aab071e9966588a8105e0edcce2736bcf3d3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:33:39.394772Z","signature_b64":"VwfRncsLwYZefl3ZeVLlO9gQa5CMt351FzLWy0jYhXiASyy7k7h8DAwHJko5g3gVqeWAEyW5pl5o69Fm3ufEAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a3bc64c8c99e7a90c4075b5fa837fb533afdd70d7198e3cf847aaafb04605626","last_reissued_at":"2026-07-05T09:33:39.394294Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:33:39.394294Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Getting By Goal Misgeneralization With a Little Help From a Mentor","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Benjamin Plaut, Mohamad H. Danesh, Nguyen X. Khanh, Tu Trinh","submitted_at":"2024-10-28T14:07:41Z","abstract_excerpt":"While reinforcement learning (RL) agents often perform well during training, they can struggle with distribution shift in real-world deployments. One particularly severe risk of distribution shift is goal misgeneralization, where the agent learns a proxy goal that coincides with the true goal during training but not during deployment. In this paper, we explore whether allowing an agent to ask for help from a supervisor in unfamiliar situations can mitigate this issue. We focus on agents trained with PPO in the CoinRun environment, a setting known to exhibit goal misgeneralization. We evaluate "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.21052","kind":"arxiv","version":3},"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/2410.21052/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":"2410.21052","created_at":"2026-07-05T09:33:39.394360+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.21052v3","created_at":"2026-07-05T09:33:39.394360+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.21052","created_at":"2026-07-05T09:33:39.394360+00:00"},{"alias_kind":"pith_short_12","alias_value":"UO6GJSGJTZ5J","created_at":"2026-07-05T09:33:39.394360+00:00"},{"alias_kind":"pith_short_16","alias_value":"UO6GJSGJTZ5JBRAH","created_at":"2026-07-05T09:33:39.394360+00:00"},{"alias_kind":"pith_short_8","alias_value":"UO6GJSGJ","created_at":"2026-07-05T09:33:39.394360+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2507.03068","citing_title":"Mitigating Goal Misgeneralization via Minimax Regret","ref_index":39,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UO6GJSGJTZ5JBRAHLNP2QN73KM","json":"https://pith.science/pith/UO6GJSGJTZ5JBRAHLNP2QN73KM.json","graph_json":"https://pith.science/api/pith-number/UO6GJSGJTZ5JBRAHLNP2QN73KM/graph.json","events_json":"https://pith.science/api/pith-number/UO6GJSGJTZ5JBRAHLNP2QN73KM/events.json","paper":"https://pith.science/paper/UO6GJSGJ"},"agent_actions":{"view_html":"https://pith.science/pith/UO6GJSGJTZ5JBRAHLNP2QN73KM","download_json":"https://pith.science/pith/UO6GJSGJTZ5JBRAHLNP2QN73KM.json","view_paper":"https://pith.science/paper/UO6GJSGJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.21052&json=true","fetch_graph":"https://pith.science/api/pith-number/UO6GJSGJTZ5JBRAHLNP2QN73KM/graph.json","fetch_events":"https://pith.science/api/pith-number/UO6GJSGJTZ5JBRAHLNP2QN73KM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UO6GJSGJTZ5JBRAHLNP2QN73KM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UO6GJSGJTZ5JBRAHLNP2QN73KM/action/storage_attestation","attest_author":"https://pith.science/pith/UO6GJSGJTZ5JBRAHLNP2QN73KM/action/author_attestation","sign_citation":"https://pith.science/pith/UO6GJSGJTZ5JBRAHLNP2QN73KM/action/citation_signature","submit_replication":"https://pith.science/pith/UO6GJSGJTZ5JBRAHLNP2QN73KM/action/replication_record"}},"created_at":"2026-07-05T09:33:39.394360+00:00","updated_at":"2026-07-05T09:33:39.394360+00:00"}