{"paper":{"title":"DriveSafe: A Framework for Risk Detection and Safety Suggestions in Driving Scenarios","license":"http://creativecommons.org/licenses/by/4.0/","headline":"DriveSafe improves driving risk assessment by conditioning it on explicit language-based scene representations.","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CV","authors_text":"Avijit Dasgupta, C. V. Jawahar, Sainithin Artham, Shankar Gangisetty","submitted_at":"2026-05-16T09:07:14Z","abstract_excerpt":"Comprehensive situational awareness is essential for autonomous vehicles operating in safety-critical environments, as it enables the identification and mitigation of potential risks. Although recent Multimodal Large Language Models (MLLMs) have shown promise on general vision-language tasks, our findings indicate that zero-shot MLLMs still underperform compared to domain-specific methods in fine-grained, spatially grounded risk assessment. To address this gap, we propose DriveSafe, a framework for risk-aware scene understanding that leverages structured natural language descriptions. Specific"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"By conditioning risk assessment on explicit language-based scene representations, DriveSafe achieves significant gains over both zero-shot MLLMs and prior domain-specific baselines. Exhaustive experiments on the DRAMA benchmark demonstrate state-of-the-art performance.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That generating spatially grounded captions enriched with multimodal context (motion, spatial, and depth cues) will provide sufficient and accurate information to enable superior risk assessment compared to direct zero-shot use of MLLMs, as stated in the abstract's motivation and method overview.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"DriveSafe improves driving risk detection by first creating detailed language-based scene descriptions enriched with motion, spatial, and depth information, then assessing risks and suggesting actions, with an adapter fine-tuned on caption-risk pairs to achieve SOTA results on the DRAMA benchmark.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"DriveSafe improves driving risk assessment by conditioning it on explicit language-based scene representations.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"8bd19c4765e92084a4b1a00d5494421d8705fa5f48a9b98b2c34c4b6367c59ca"},"source":{"id":"2605.16892","kind":"arxiv","version":1},"verdict":{"id":"80ae91de-ecc4-459e-8a08-68497ce470fd","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T21:36:27.014475Z","strongest_claim":"By conditioning risk assessment on explicit language-based scene representations, DriveSafe achieves significant gains over both zero-shot MLLMs and prior domain-specific baselines. Exhaustive experiments on the DRAMA benchmark demonstrate state-of-the-art performance.","one_line_summary":"DriveSafe improves driving risk detection by first creating detailed language-based scene descriptions enriched with motion, spatial, and depth information, then assessing risks and suggesting actions, with an adapter fine-tuned on caption-risk pairs to achieve SOTA results on the DRAMA benchmark.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That generating spatially grounded captions enriched with multimodal context (motion, spatial, and depth cues) will provide sufficient and accurate information to enable superior risk assessment compared to direct zero-shot use of MLLMs, as stated in the abstract's motivation and method overview.","pith_extraction_headline":"DriveSafe improves driving risk assessment by conditioning it on explicit language-based scene representations."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.16892/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T22:01:19.544665Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T21:50:54.901400Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"cited_work_retraction","ran_at":"2026-05-19T20:52:41.618454Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T18:41:56.282464Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T18:33:26.360916Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"361f440c94ddb65f7fd8d9f0cd90772f5ebd047455ca8f0233ecf3fdd6ca68aa"},"references":{"count":44,"sample":[{"doi":"","year":2015,"title":"Risk assessment, modelling and proactive safety manage- ment system in aviation: a literature review,","work_id":"a99bf928-949c-460f-ac84-e6e94abc8cf4","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"Risk management in the healthcare safety management system,","work_id":"294d6a14-476c-46cc-a3e9-eb2c6a556eaa","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2019,"title":"Safety assessment of collaborative robotics through automated formal verifi- cation,","work_id":"08bff5a6-2050-4d81-a256-3a87d80aa60f","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Road traffic injuries fact sheet,","work_id":"89b2feab-8296-4bc0-aec9-b6c6b74c93f3","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Fatality statistics: State-by- state,","work_id":"89a636f8-aaa4-42bc-ae29-1e5cc38d10fa","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":44,"snapshot_sha256":"6a6d921ae9e47632f219c0b008fd443560f2cb8f5207f4f1009d794abe5c1dcc","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"}