AI Paper Atlas

Research map / Architectures and efficiency / Quantization

RaBiT: Residual Aware Binarization Training for Accurate and Efficient LLMs

Youngcheon You, Banseok Lee, Minseop Choi, Seonyoung Kim, Hyochan Chong, Changdong Kim, Youngmin Kim, Dongkyu Kim

ICML 2026 · Architectures and efficiency › Quantization · approach: Post-training quantization of LLMs · 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
use coupled QAT (shared FP weight, sequential residual paths) instead of standard QAT (ablation)RaBiT 2-bit · WikiText-2, Llama2-7Bperplexity: 6.55 -> 5.84Improved
add iterative residual SVID and importance-scaled preconditioning to coupled QAT (ablation)RaBiT · WikiText-2, Llama2-7Bperplexity: 5.84 -> 5.78Improved
add iterative SVID and preconditioning to standard QAT (ablation)Standard QAT · WikiText-2, Llama2-7Bperplexity: 6.55 -> 6.18Improved
replace greedy SVID with iterative residual SVID init (ablation)RaBiT init · Initial KL loss, Llama2-7B q_projKL loss: 17,152 -> 13,760Improved
add I/O channel importance-scaled preconditioning to init (ablation)RaBiT init · Initial KL loss, Llama2-7B q_projKL loss: 13,760 -> 2,672Improved
add I/O channel importance scaling (weight reconstruction view) (ablation)RaBiT init · Weight reconstruction, Llama2-7B q_projAvg MSE: 0.122 -> 0.302Hurt
use coupled residual training vs standard QAT on path behavior (ablation)RaBiT · Llama2-7B layer 5 residual alignmentCorr(R1,y2): 0.4395 -> 0.6484Improved
freeze primary binary core (MBOK-style) instead of coupled training (ablation)RaBiT variant · Llama2-7B training dynamicsinter-path correlation / KL lossHurt
freeze scales or binary cores in coupled training (ablation)RaBiT variants (Scale-only, Scale-frozen) · Llama2-7B training lossKL lossHurt
use RaBiT instead of QTIP at 2 bitsRaBiT vs QTIP · WikiText-2, Llama2-7Bperplexity: 5.86 -> 5.78Improved
use RaBiT instead of QTIP at 2 bitsRaBiT vs QTIP · WikiText-2, Llama2-13Bperplexity: 5.11 -> 5.15Hurt
use RaBiT instead of QTIP on hard benchmarksRaBiT vs QTIP · Challenging benchmarks avg, Llama2-13Baccuracy: 25.38 -> 27.14Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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