{"paper":{"title":"Seeing the Scene Matters: Revealing Forgetting in Video Understanding Models with a Scene-Aware Long-Video Benchmark","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"Vision-language models forget long-range scene context in videos, shown by a new benchmark with sharp accuracy drops.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chao Li, Chenglam Ho, Hao Chen, Jinping Wang, Seng Nam Chen, Xinyu Mao, Yu Zhang","submitted_at":"2026-03-28T12:44:19Z","abstract_excerpt":"Long video understanding (LVU) remains a core challenge in multimodal learning. Although recent vision-language models (VLMs) have made notable progress, existing benchmarks mainly focus on either fine-grained perception or coarse summarization, offering limited insight into temporal understanding over long contexts. In this work, we define a scene as a coherent segment of a video in which both visual and semantic contexts remain consistent, aligning with human perception. This leads us to a key question: can current VLMs reason effectively over long, scene-level contexts? To answer this, we i"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Our evaluation reveals a sharp drop in accuracy when VLMs attempt to answer scene-level questions, indicating significant forgetting of long-range context.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the authors' definition of a scene as a coherent segment with consistent visual and semantic contexts accurately isolates long-range forgetting, and that the benchmark questions do not introduce other confounds in video selection or question design.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SceneBench shows VLMs lose accuracy on scene-level questions in long videos due to forgetting, and Scene-RAG retrieval improves performance by 2.5%.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Vision-language models forget long-range scene context in videos, shown by a new benchmark with sharp accuracy drops.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"a917d5f8448758645675e74b0e4a912d18ccb4bfee6a590719933b28662f355d"},"source":{"id":"2603.27259","kind":"arxiv","version":3},"verdict":{"id":"b363ad80-56a4-4b57-bcbe-4a1a8a10ddcb","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T22:02:10.306737Z","strongest_claim":"Our evaluation reveals a sharp drop in accuracy when VLMs attempt to answer scene-level questions, indicating significant forgetting of long-range context.","one_line_summary":"SceneBench shows VLMs lose accuracy on scene-level questions in long videos due to forgetting, and Scene-RAG retrieval improves performance by 2.5%.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the authors' definition of a scene as a coherent segment with consistent visual and semantic contexts accurately isolates long-range forgetting, and that the benchmark questions do not introduce other confounds in video selection or question design.","pith_extraction_headline":"Vision-language models forget long-range scene context in videos, shown by a new benchmark with sharp accuracy drops."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2603.27259/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"}