Metraipolis › archive › dc_412c90285efd
Activation Outliers Matter: Robust Recovery for Quantized Multimodal LLMs — arXiv:2608.26581
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.26581
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T16:09:43.049Z
- SHA-256 of the archived text
01db7d51299d1c2627f18e61f4fd33087eb871f913c63632998d4b51e382a5fd- Stable URL of this page
https://metraipolis.com/documento/dc_412c90285efd
How to cite
Activation Outliers Matter: Robust Recovery for Quantized Multimodal LLMs — arXiv:2608.26581 — arXiv (en), https://arxiv.org/abs/2608.26581 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_412c90285efd (Metraipolis archive dc_412c90285efd, sha256 01db7d51299d1c2627f18e61f4fd33087eb871f913c63632998d4b51e382a5fd)
BibTeX
@misc{dc412c90285efd,
title = {Activation Outliers Matter: Robust Recovery for Quantized Multimodal LLMs — arXiv:2608.26581},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.26581},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_412c90285efd; sha256 01db7d51299d1c2627f18e61f4fd33087eb871f913c63632998d4b51e382a5fd},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.26581. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Low-bit quantization offers a promising avenue for reducing the computational and memory demands of Multimodal Large Language Models (MLLMs). Recent hardware support for low-precision formats, ranging from MXFP8 to ultra-low-bit formats such as MXFP4 and HiF4, has accelerated research into efficient MLLM training and deployment. In this work, we present a systematic study of these quantization schemes in representative MLLMs that span both video generation and reasoning tasks. Our analysis shows that MXFP8 achieves near-lossless performance, whereas aggressive 4-bit quantization leads to significant degradation. Through extensive ablations, we identify activation quantization as the primary source of this performance loss, contributing substantially more than weight quantization. Motivated by this observation, we propose Residual Fallback Quantization (RFQ), a lightweight activation reconstruction framework that supplements the primary ulta-low-bit activation representation with an auxiliary quantized residual pathway. By explicitly modeling and compensating for quantization errors, RFQ improves activation fidelity while preserving the efficiency advantages of ultra-low-bit computation. RFQ requires no architectural modifications and incurs negligible computational overhead. Extensive experiments on Wan2.2 and Qwen3-VL demonstrate that RFQ consistently recovers a substantial portion of the performance lost under the quantization of MXFP4 and HiF4, significantly narrowing the gap to BF16 baselines across both generation and 4 reasoning benchmarks. Our findings establish activation quantization as the dominant bottleneck in ultra-low-bit MLLMs and highlight residual-based activation reconstruction as an effective and practical strategy for robust 4-bit deployment. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-28 in cs.LG). Categorie arXiv: Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2608.26581 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-17T15:34:49.911Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-08-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).