{"paper":{"title":"Learning from Trials and Errors: Reflective Test-Time Planning for Embodied LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Embodied LLMs improve long-horizon task performance by reflecting on failures before and after each execution at test time.","cross_cats":["cs.AI","cs.CL","cs.CV","cs.RO"],"primary_cat":"cs.LG","authors_text":"Huang Huang, Jiajun Wu, Leonidas Guibas, Li Fei-Fei, Manling Li, Yejin Choi, Yining Hong","submitted_at":"2026-02-24T18:55:18Z","abstract_excerpt":"Embodied LLMs endow robots with high-level task reasoning, but they cannot reflect on what went wrong or why, turning deployment into a sequence of independent trials where mistakes repeat rather than accumulate into experience. Drawing upon human reflective practitioners, we introduce Reflective Test-Time Planning, which integrates two modes of reflection: \\textit{reflection-in-action}, where the agent uses test-time scaling to generate and score multiple candidate actions using internal reflections before execution; and \\textit{reflection-on-action}, which uses test-time training to update b"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments on our newly-designed Long-Horizon Household benchmark and MuJoCo Cupboard Fitting benchmark show significant gains over baseline models, with zero-shot generalization to photorealistic HM3D environments and real-robot experiments on a Franka Panda arm. Ablations confirm that reflection-in-action and reflection-on-action are mutually dependent, and that retrospective reflection achieves better credit assignment than step-wise external feedback at lower computational overhead.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That internal model reflections and external feedback after execution can reliably identify the causes of failures and that test-time training updates can improve the policy without instability or loss of prior capabilities.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Reflective Test-Time Planning combines pre-execution internal reflection with post-execution model updates to improve embodied LLMs on household and manipulation tasks with better long-horizon credit assignment.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Embodied LLMs improve long-horizon task performance by reflecting on failures before and after each execution at test time.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"6aeef0b29f34d89eca50c8277f822abc0fecc255f08d3a7ee00447d702b814fe"},"source":{"id":"2602.21198","kind":"arxiv","version":3},"verdict":{"id":"76727724-db0d-4b3e-a149-bf9ab9226bd3","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T19:41:10.555564Z","strongest_claim":"Experiments on our newly-designed Long-Horizon Household benchmark and MuJoCo Cupboard Fitting benchmark show significant gains over baseline models, with zero-shot generalization to photorealistic HM3D environments and real-robot experiments on a Franka Panda arm. Ablations confirm that reflection-in-action and reflection-on-action are mutually dependent, and that retrospective reflection achieves better credit assignment than step-wise external feedback at lower computational overhead.","one_line_summary":"Reflective Test-Time Planning combines pre-execution internal reflection with post-execution model updates to improve embodied LLMs on household and manipulation tasks with better long-horizon credit assignment.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That internal model reflections and external feedback after execution can reliably identify the causes of failures and that test-time training updates can improve the policy without instability or loss of prior capabilities.","pith_extraction_headline":"Embodied LLMs improve long-horizon task performance by reflecting on failures before and after each execution at test time."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2602.21198/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":2,"snapshot_sha256":"122c40ea650ad326a3f0460910b38b598ec58449d188a4df0cb8e6f99d1b497a"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}