AI Paper Atlas

Research map / Benchmarks / TempCompass

TempCompass leaderboard

19 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on TempCompass. Best accuracy: 77.2 (SpecTemp (47.6 frames avg), CVPR 2026).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

TempCompass · accuracy

19 papers, each with its best reported number for its own method; higher is better.

#accuracy ↑MethodPaperTopic
177.2SpecTemp (47.6 frames avg)
Qwen2.5-VL-7B target + 3B draft
Thinking with Drafts: Speculative Temporal Reasoning for Efficient Long Video Understanding
CVPR 2026 · Table 1
Video-language understanding
276.76DeltaVid-Qwen3-VL-8B
Qwen3-VL-8B
DELTAVID: Enhancing Fine-Grained Spatiotemporal Perception with Cross-Video Differences
NeurIPS 2026 · Table 2
Video-language understanding
376.2RLER
Qwen2.5-VL-7B
Reinforce to Learn, Elect to Reason: A Dual Paradigm for Video Reasoning
CVPR 2026 · Table 1
Video-language understanding
475.9AdaCodec (comparable token budget)
Qwen3-VL-8B
AdaCodec: A Predictive Visual Code for Video MLLMs
NeurIPS 2026 · Table 1
Video-language understanding
575.6Video-R1 + FFR
Video-R1-SFT 7B
Find, Fix, Reason: Context Repair for Video Reasoning
ICML 2026 · Table 2
Reasoning and chain of thought
675.1Qwen3-VL-8B + MotionAtlas-Data
Qwen3-VL-8B
MotionAtlas: A High-Quality Dataset and Benchmark for Dense Motion Captioning
ECCV 2026 · Table 6
Video generation
774Video-VER
Qwen2.5-VL-7B
When Thinking Drifts: Evidential Grounding for Robust Video Reasoning
NeurIPS 2025 · Table 1
Multimodal LLMs
873.7VideoRFT
Qwen2.5-VL-7B
VideoRFT: Incentivizing Video Reasoning Capability in MLLMs via Reinforced Fine-Tuning
NeurIPS 2025 · Performance Comparison table
Instruction and fine-tuning
973.5Qwen2.5-VL-7B + CH-M
Qwen2.5-VL-7B
Unstitching the Chimera: Frame-Level Risk and Train-Free Mitigation for Video Hallucination
CVPR 2026 · Table 1
Video-language understanding
1073.5Video-KTR
Qwen2.5-VL-7B (64 frames)
Video-KTR: Reinforcing Video Reasoning via Key Token Attribution
ICLR 2026 · Table 1
Video-language understanding
9 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.