AI Paper Atlas

Research map / Video / Video-language understanding

From Content to Knowledge: Lightning Fast Long-Video Understanding with Neural Knowledge Representations

Yuchen Guan, Xiao Li, Zongyu Guo, Xiaoyi Zhang, Xiulian Peng, Chun Yuan, Yan Lu

ICML 2026 · Video › Video-language understanding · approach: Long video understanding · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
upgrade NKR backbone from Qwen3-8B to Qwen3-14B (ablation)NKR · LVBenchoverall accuracy: 39.5 -> 48.8 (speed 0.26 -> 0.33 s/query)Improved
raise LoRA rank from 64 to 256 (ablation)NKR on Qwen3-14B · LVBenchoverall accuracy: 44.2 -> 48.8Improved
add clip-level and video-level QA to dense descriptions (ablation)NKR training data · LVBenchoverall accuracy: 14.3 -> 48.8Improved
add dense descriptions to QA data (ablation)NKR training data · LVBenchoverall accuracy: 38.8 -> 48.8Improved
add video-level QA (ablation)NKR training data · LVBenchoverall accuracy: 45.0 -> 48.8Improved
replace direct video tokens with NKR weightsVLM-based long video QA · LVBenchinference time per query: 31.7s-64.5s -> 0.33sImproved
use NKR instead of video tokens for videos longer than one hourNKR · LVBench >1h vs <1haccuracy drop: -0.3 (NKR) vs -6.7 (GPT4.1)Improved
use NKR instead of retrieval agent VideoRAGNKR vs VideoRAG · LVBench reasoning questionsaccuracy: 43.9 -> 47.1Improved
use NKR instead of retrieval agent VideoRAGNKR vs VideoRAG · LVBench key information retrievalaccuracy: 57.1 -> 53.0Hurt
apply NKR to images with Qwen2.5-VL-7BQwen2.5-VL-7B · MME-RealWorld subsetoverall accuracy: 32.1 -> 35.4Improved
apply NKR to images with Qwen2.5-VL-7BQwen2.5-VL-7B · MME-RealWorld subset perceptionaccuracy: 39.9 -> 38.4Hurt
replace NKR with Deep Video Discovery agentNKR · LVBenchoverall accuracy: 48.8 -> 74.2 (0.33 -> 180.0 s/query)Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy