{"paper":{"title":"VECTOR-Drive: Tightly Coupled Vision-Language and Trajectory Expert Routing for End-to-End Autonomous Driving","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"By routing tokens to separate vision-language and trajectory experts while sharing self-attention, VECTOR-DRIVE resolves the coupling trade-off in vision-language-action models for autonomous driving.","cross_cats":["cs.AI","cs.RO"],"primary_cat":"cs.CV","authors_text":"Fei Gao, Jianlin Yu, Jiaqiao Liu, Rui Zhao, Zhenhai Gao","submitted_at":"2026-05-09T09:34:50Z","abstract_excerpt":"End-to-end autonomous driving requires models to understand traffic scenes, infer driving intent, and generate executable motion plans. Recent vision-language-action (VLA) models inherit semantic priors from large-scale vision-language pretraining, yet still face a coupling trade-off: fully shared backbones preserve multimodal interaction but may entangle language reasoning and trajectory prediction, whereas decou pled reasoning-action pipelines reduce task conflict but weaken semantic-motion coupling. We propose VECTOR-DRIVE, a tightly coupled VLA framework built on Qwen2.5-VL-3B. VECTOR-DRIV"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On Bench2Drive, VECTOR-DRIVE achieves 88.91 Driving Score and outperforms representative end-to-end and VLA-based baselines.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That routing tokens to separate experts according to semantics while keeping shared self-attention sufficiently preserves multimodal coupling without introducing new task conflicts or information loss.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"VECTOR-DRIVE couples vision-language reasoning and trajectory planning in one Transformer via semantic expert routing and flow-matching, reaching 88.91 driving score on Bench2Drive.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"By routing tokens to separate vision-language and trajectory experts while sharing self-attention, VECTOR-DRIVE resolves the coupling trade-off in vision-language-action models for autonomous driving.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"485b5f85a72d212b78423e34c7d40ffb1cdef328b515c3e7abc17634331c72f9"},"source":{"id":"2605.08830","kind":"arxiv","version":2},"verdict":{"id":"fe341ff8-a9a3-41f2-98ba-01cd71cb830d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-12T01:44:57.024653Z","strongest_claim":"On Bench2Drive, VECTOR-DRIVE achieves 88.91 Driving Score and outperforms representative end-to-end and VLA-based baselines.","one_line_summary":"VECTOR-DRIVE couples vision-language reasoning and trajectory planning in one Transformer via semantic expert routing and flow-matching, reaching 88.91 driving score on Bench2Drive.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That routing tokens to separate experts according to semantics while keeping shared self-attention sufficiently preserves multimodal coupling without introducing new task conflicts or information loss.","pith_extraction_headline":"By routing tokens to separate vision-language and trajectory experts while sharing self-attention, VECTOR-DRIVE resolves the coupling trade-off in vision-language-action models for autonomous driving."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.08830/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-19T22:34:02.009475Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T14:01:21.834211Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T10:47:58.717701Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"8bedaf2e714cbc13af518ded835eb8ec1736b6191913fc6513f9bba188f129a4"},"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"}