AI Paper Atlas

Research map / Architectures and efficiency / Quantization

SPEED-Q: Staged Processing with Enhanced Distillation Towards Efficient Low-Bit On-Device VLM Quantization

Tianyu Guo, Shanwei Zhao, Shiai Zhu, Chenguang Ma

AAAI 2026 · Architectures and efficiency › Quantization · approach: Post-training quantization of ViTs · 5 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
add distillation loss to task loss (ablation)SPEED-Q (InternVL3-1B, 4-bit) · MMStaraccuracy: 49.33 -> 51.07Improved
add distillation loss to task loss (ablation)SPEED-Q (InternVL3-1B, 4-bit) · AI2Daccuracy: 66.13 -> 68.81Improved
add distillation loss to task loss (ablation)SPEED-Q (InternVL3-1B, 4-bit) · OCRBenchaccuracy: 75.5 -> 79.7Improved
add task loss to distillation loss (ablation)SPEED-Q (InternVL3-1B, 4-bit) · MMStaraccuracy: 50.40 -> 51.07Improved
add task loss to distillation loss (ablation)SPEED-Q (InternVL3-1B, 4-bit) · AI2Daccuracy: 68.46 -> 68.81Improved
add task loss to distillation loss (ablation)SPEED-Q (InternVL3-1B, 4-bit) · OCRBenchaccuracy: 77.9 -> 79.7Improved
use staged instead of joint quantization (ablation)SPEED-Q (InternVL3-1B, 4-bit) · MMStaraccuracy: 49.67 -> 51.07Improved
use staged instead of joint quantization (ablation)SPEED-Q (InternVL3-1B, 4-bit) · AI2Daccuracy: 68.46 -> 68.81Improved
use staged instead of joint quantization (ablation)SPEED-Q (InternVL3-1B, 4-bit) · OCRBenchaccuracy: 77.0 -> 79.7Improved
SPEED-Q instead of MBQ at 2-bitInternVL3-1B · Avg. of 8 benchmarks (2-bit)average accuracy: 3.81 -> 51.48Improved
SPEED-Q instead of MBQ at 4-bitInternVL3-8B · Avg. of 8 benchmarks (4-bit)average accuracy: 75.35 -> 75.04No clear effect
SPEED-Q instead of MBQ at 2-bitInternVL3-8B · Avg. of 8 benchmarks (2-bit)average accuracy: 61.66 -> 68.46Improved

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