{"paper":{"title":"Contextual Multi-Task Reinforcement Learning for Autonomous Reef Monitoring","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Contextual multi-task reinforcement learning trains one policy to handle multiple reef monitoring tasks.","cross_cats":["cs.AI"],"primary_cat":"cs.RO","authors_text":"Frank Kirchner, Mariela De Lucas Alvarez, Melvin Laux, Rebecca Adam, Rina Alo, S\\\"oren T\\\"opper, Yi-Ling Liu","submitted_at":"2026-04-14T12:16:56Z","abstract_excerpt":"Although autonomous underwater vehicles promise the capability of marine ecosystem monitoring, their deployment is fundamentally limited by the difficulty of controlling vehicles under highly uncertain and non-stationary underwater dynamics. To address these challenges, we employ a data-driven reinforcement learning approach to compensate for unknown dynamics and task variations. Traditional single-task reinforcement learning has a tendency to overfit the training environment, thus, limit the long-term usefulness of the learnt policy. Hence, we propose to use a contextual multi-task reinforcem"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We train a single context-dependent policy that is able to solve multiple related monitoring tasks in a simulated reef environment in HoloOcean.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That performance gains observed in the HoloOcean simulator will translate to real-world underwater environments with highly uncertain and non-stationary dynamics.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"A context-dependent multi-task RL policy is trained and evaluated in HoloOcean simulation to solve multiple reef monitoring tasks with claimed improvements in sample efficiency, zero-shot generalization, and robustness to water currents.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Contextual multi-task reinforcement learning trains one policy to handle multiple reef monitoring tasks.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"d0fcdf051c9c3faa3c6730792595ead42e55e923da379ff3e2d053957aedbf4f"},"source":{"id":"2604.12645","kind":"arxiv","version":2},"verdict":{"id":"7cc8bb5f-551b-4202-a9da-37d9946a83cb","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T15:08:54.040768Z","strongest_claim":"We train a single context-dependent policy that is able to solve multiple related monitoring tasks in a simulated reef environment in HoloOcean.","one_line_summary":"A context-dependent multi-task RL policy is trained and evaluated in HoloOcean simulation to solve multiple reef monitoring tasks with claimed improvements in sample efficiency, zero-shot generalization, and robustness to water currents.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That performance gains observed in the HoloOcean simulator will translate to real-world underwater environments with highly uncertain and non-stationary dynamics.","pith_extraction_headline":"Contextual multi-task reinforcement learning trains one policy to handle multiple reef monitoring tasks."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.12645/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"}