Research map / Video / Video-language understanding
NeuS-QA: Grounding Long-Form Video Understanding in Temporal Logic and Neuro-Symbolic Reasoning
Sahil Shah, S P Sharan, Harsh Goel, Minkyu Choi, Mustafa Munir, Manvik Pasula, Radu Marculescu, Sandeep Chinchali
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add NeuS-QA segment retrieval to Qwen2.5-VL | Qwen2.5-VL-7B · LongVideoBench (reasoning subset) | overall accuracy (%): 50.44 -> 60.09 | Improved |
| add NeuS-QA to GPT-4o | GPT-4o · LongVideoBench (reasoning subset) | overall accuracy (%): 49.56 -> 58.33 | Improved |
| add NeuS-QA to LLaVA-OneVision | LLaVA-OneVision-Qwen2-7B · LongVideoBench (reasoning subset) | overall accuracy (%): 53.07 -> 58.77 | Improved |
| add NeuS-QA to Aria | Aria · LongVideoBench (reasoning subset) | overall accuracy (%): 53.07 -> 58.77 | Improved |
| add NeuS-QA to Qwen2.5-VL on CinePile | Qwen2.5-VL-7B · CinePile | accuracy (%): 50.73 -> 53.66 | Improved |
| add NeuS-QA to GPT-4o on E3E questions | GPT-4o · LongVideoBench E3E | accuracy (%): 58.11 -> 68.92 | Improved |
| add NeuS-QA to GPT-4o on T3E questions | GPT-4o · LongVideoBench T3E | accuracy (%): 43.40 -> 49.06 | Mixed |
| replace temporal-logic automaton with VLM-only relevance judgement (ablation) | NeuS-QA · LongVideoBench (GPT-4o answerer, ablation) | accuracy (%): 40.09 -> 58.33 | Improved |
| use InternVL2-8B instead of Qwen2.5-VL to build the automaton (ablation) | NeuS-QA · LongVideoBench (GPT-4o answerer, ablation) | accuracy (%): 52.63 -> 58.33 | Improved |
| trim to relevant segments instead of dynamic scaling (ablation) | NeuS-QA · LongVideoBench (GPT-4o answerer, ablation) | accuracy (%): 53.51 -> 58.33 | Improved |
| trim to relevant segments instead of speed bias sampling (ablation) | NeuS-QA · LongVideoBench (GPT-4o answerer, ablation) | accuracy (%): 52.19 -> 58.33 | Improved |
| use NeuS-QA segments instead of ground-truth annotated segments | NeuS-QA · LongVideoBench | accuracy (%): 52 -> 59 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- VideoTree — VideoTree: Adaptive Tree-based Video Representation for LLM Reasoning on Long Videos (CVPR 2025) · better on 1 of 1 shared result
- VideoAgent — VideoAgent: A Memory-augmented Multimodal Agent for Video Understanding (ECCV 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding (NeurIPS 2024 · similarity 0.91)
- Temporal Chain of Thought: Long-Video Understanding by Thinking in Frames (NeurIPS 2025 · similarity 0.90)
- VideoEspresso: A Large-Scale Chain-of-Thought Dataset for Fine-Grained Video Reasoning via Core Frame Selection (CVPR 2025 · similarity 0.90)
- Not Another Text Benchmark: Putting the “Visual" Back in Visual Question Answering for Large Video Models (NeurIPS 2026 · similarity 0.90)
- Streaming Long Video Understanding with Large Language Models (NeurIPS 2024 · similarity 0.89)
- Enhancing Video-LLM Reasoning via Agent-of-Thoughts Distillation (CVPR 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
