{"paper":{"title":"Learning Without Losing Identity: Capability Evolution for Embodied Agents","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Embodied agents improve task success from 32% to 91% by evolving separate capability modules without altering their core identity or safety limits.","cross_cats":["cs.AI"],"primary_cat":"cs.RO","authors_text":"Cong Yang, John See, Simin Luan, Xue Qin, Zhijun Li","submitted_at":"2026-04-09T04:51:07Z","abstract_excerpt":"Embodied agents are expected to operate persistently in dynamic physical environments, continuously acquiring new capabilities over time. Existing approaches to improving agent performance often rely on modifying the agent itself -- through prompt engineering, policy updates, or structural redesign -- leading to instability and loss of identity in long-lived systems. In this work, we propose a capability-centric evolution paradigm for embodied agents. We argue that a robot should maintain a persistent agent as its cognitive identity, while enabling continuous improvement through the evolution "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We demonstrate through simulated embodied tasks that capability evolution improves task success rates from 32.4% to 91.3% over 20 iterations, outperforming both agent-modification baselines and established skill-learning methods (SPiRL, SkiMo), while preserving zero policy drift and zero safety violations.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That embodied capabilities can be cleanly decomposed into independent, versioned ECMs whose evolution leaves the persistent agent identity and safety constraints untouched.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Embodied agents maintain a persistent identity while evolving capabilities via modular ECMs, raising simulated task success from 32.4% to 91.3% over 20 iterations with zero policy drift or safety violations.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Embodied agents improve task success from 32% to 91% by evolving separate capability modules without altering their core identity or safety limits.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"025bce8f149746d01a17d1f66eaeb298cd6927cbe705085d74d200b95cc9f585"},"source":{"id":"2604.07799","kind":"arxiv","version":2},"verdict":{"id":"2b3a9046-31dd-4f3a-b9fa-e8c8f559aa4c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T18:05:59.766335Z","strongest_claim":"We demonstrate through simulated embodied tasks that capability evolution improves task success rates from 32.4% to 91.3% over 20 iterations, outperforming both agent-modification baselines and established skill-learning methods (SPiRL, SkiMo), while preserving zero policy drift and zero safety violations.","one_line_summary":"Embodied agents maintain a persistent identity while evolving capabilities via modular ECMs, raising simulated task success from 32.4% to 91.3% over 20 iterations with zero policy drift or safety violations.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That embodied capabilities can be cleanly decomposed into independent, versioned ECMs whose evolution leaves the persistent agent identity and safety constraints untouched.","pith_extraction_headline":"Embodied agents improve task success from 32% to 91% by evolving separate capability modules without altering their core identity or safety limits."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.07799/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"}