{"paper":{"title":"VISOR: A Vision-Language Model-based Test Oracle for Testing Robots","license":"http://creativecommons.org/licenses/by/4.0/","headline":"VISOR uses vision-language models to automatically score robot task correctness, quality, and uncertainty from videos.","cross_cats":["cs.RO"],"primary_cat":"cs.SE","authors_text":"Aitor Arrieta, Pablo Valle, Paolo Arcaini, Prasun Saurabh, Shaukat Ali","submitted_at":"2026-05-11T11:46:57Z","abstract_excerpt":"Testing robots requires assessing whether they perform their intended tasks correctly, dependably, and with high quality, a challenge known as the test oracle problem in software testing. Traditionally, this assessment relies on task-specific symbolic oracles for task correctness and on human manual evaluation of robot behavior, which is time-consuming, subjective, and error-prone. To address this, we propose VISOR, a Vision-Language Model (VLM)-based approach for automated test oracle assessment that eliminates the need of expensive human evaluations. VISOR performs automated evaluation of ta"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"VISOR performs automated evaluation of task correctness and quality, addressing the limitations of existing symbolic test oracles, which are task-specific and provide pass/fail judgments without explicitly quantifying task quality. Given the inherent uncertainty in VLMs, VISOR also explicitly quantifies its own uncertainty during test assessments.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That off-the-shelf vision-language models can accurately interpret and score complex, dynamic robot behaviors in videos without task-specific fine-tuning or symbolic grounding.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"VISOR applies VLMs to automate robot test oracles for correctness and quality assessment while reporting uncertainty, with evaluation on GPT and Gemini showing trade-offs in precision and recall but poor uncertainty calibration.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"VISOR uses vision-language models to automatically score robot task correctness, quality, and uncertainty from videos.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"0f5bcdb1d242da8db8a1aea28a207eacab06badbdeacefceed3440a8d0ee0ef9"},"source":{"id":"2605.10408","kind":"arxiv","version":2},"verdict":{"id":"6dab8467-76c3-4365-9f96-55e50c56b1cf","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-12T05:13:20.417354Z","strongest_claim":"VISOR performs automated evaluation of task correctness and quality, addressing the limitations of existing symbolic test oracles, which are task-specific and provide pass/fail judgments without explicitly quantifying task quality. Given the inherent uncertainty in VLMs, VISOR also explicitly quantifies its own uncertainty during test assessments.","one_line_summary":"VISOR applies VLMs to automate robot test oracles for correctness and quality assessment while reporting uncertainty, with evaluation on GPT and Gemini showing trade-offs in precision and recall but poor uncertainty calibration.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That off-the-shelf vision-language models can accurately interpret and score complex, dynamic robot behaviors in videos without task-specific fine-tuning or symbolic grounding.","pith_extraction_headline":"VISOR uses vision-language models to automatically score robot task correctness, quality, and uncertainty from videos."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.10408/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-19T15:34:08.894681Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T11:31:18.101957Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T09:22:33.475475Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"00eb4d315217eab646312c45e8a3ee052af60aab2f428f7336b3a3ac98248d49"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":3,"snapshot_sha256":"abfe402594a550b644f7fb29f18e0e3cf86886ff208985794256136ff1216a74"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}