# EMCompress: Video-LLMs with Endomorphic Multimodal Compression

Source: [EMCompress: Video-LLMs with Endomorphic Multimodal Compression](<https://www.sota2.com/research/paper/68fa7b571e209131cea87eaf>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-29T05:00:05.003Z

Authors: Zheyu Fan, Jiateng Liu, Yuji Zhang, Zihan Wang, Yi R. Fung, Manling Li, Heng Ji

Published: 2025-08-27T14:33:32.000Z

## Abstract

Video-LLMs face a fundamental tension in long-video reasoning: static, sparse frame sampling either dilutes evidence across task-irrelevant segments at significant cost or misses fine-grained temporal semantics altogether. We propose a novel, cognitively-inspired task -- Endomorphic Multimodal Compression (EMC) -- as a structurally-constrained sufficient-statistic problem for VideoQA, and formulate it as an endomorphic transformation F\_EMC : (V, Q) -\> (v, q) that compresses the multimodal input while preserving answer invariance across reasonable downstream models. The endomorphic form keeps the compressed output in the downstream pipeline's native task space -- a structural mirror of the filter-then-reason mechanism in the cognitive literature motivating EMC -- distinguishing it from latent-code compression (IB / VIB) and making the formulation extensible to other multimodal settings. Under the Markov chain A -\> (V, Q) -\> (v, q), EMC realizes the classical sufficiency condition I((v, q); A) = I((V, Q); A) in its VideoQA-natural form. As a modular front-end, EMC plugs into both Video Instruction Tuning and Video Question Answering pipelines. We release the first dedicated benchmark and propose ReSimplifyIt, an EMC baseline surpassing prior methods by 0.40 F-1 with competitive query rewriting. Integrating EMC yields relative gains of 7.33% in training and 33.7% in inference for video-language understanding.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Video Question Answering on EMCompress (Stage-1)](<https://www.sota2.com/research/sota/video-question-answering-on-emcompress-stage-1.md>): Video Question Answering; EMCompress (Stage-1); Temporal Relational mIoU: 23.

- [Video Temporal Grounding on EMCompress 1.0 (test)](<https://www.sota2.com/research/sota/video-temporal-grounding-on-emcompress-1-0-test.md>): Video Temporal Grounding; EMCompress 1.0 (test); Temporal Relational mIoU: 23.

- [Query Rewriting on EMCompress](<https://www.sota2.com/research/sota/query-rewriting-on-emcompress.md>): Query Rewriting; EMCompress; Temporal Relational Score: 66.8.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2508.21094>)

[PDF](<https://arxiv.org/pdf/2508.21094.pdf>)
