AI Paper Atlas

Research map / Architectures and efficiency / Quantization

TetraJet-v2: Accurate NVFP4 Training for Large Language Models with Oscillation Suppression and Outlier Control

Yuxiang Chen, Yifan Liu, Xiaoming Xu, Pengle Zhang, Michael Beyer, Martin Rapp, Jun Zhu, Jianfei Chen

ICML 2026 · Architectures and efficiency › Quantization · approach: Ultra-low-bit LLM quantization · 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
add OsciReset (reset oscillating master weights to bin centers) (ablation)TetraJet-v2-base+OC · OLMo2-150M, 107B tokensvalidation perplexityImproved
add Q-EMA, Freeze or Dampen oscillation suppression instead of OsciReset (ablation)TetraJet-v2-base+OC · OLMo2-150M, 107B tokensvalidation perplexityHurt
add OutControl (retain top 10% activation outlier channels in higher precision) (ablation)TetraJet-v2-base · OLMo2-370M, 212B tokensvalidation perplexityImproved
add OsciReset and OutControl to TetraJet-v2-base (ablation)TetraJet-v2-base · OLMo2-370M, C4 validationvalidation perplexity: 25.50 -> 25.11Improved
add OsciReset and OutControl to TetraJet-v2-base (ablation)TetraJet-v2-base · OLMo2-70M, C4 validationvalidation perplexity: 49.33 -> 47.75Improved
apply Random Hadamard Transform in the forward pass (ablation)NVFP4 linear layer · OLMo2 NVFP4 trainingoptimization qualityHurt
apply RHT to both dX and dW backward matmuls (ablation)NVFP4 linear layer · OLMo2 NVFP4 trainingoptimization qualityImproved
quantize both backward matmuls rather than one (ablation)NVFP4 linear layer · OLMo2-370M, 52B tokensvalidation perplexity delta vs BF16: -0.01 and +0.07 individually -> +0.21 combinedHurt
keep mlp.ffn2 in BF16 (ablation)fully quantized NVFP4 model · OLMo2-370M, 52B tokensvalidation perplexity: 24.89 -> 24.54Improved
switch to FP6 x FP4 for the last 50B tokensTetraJet-v2-full · OLMo2-370M, 212B tokensvalidation perplexityImproved
use stochastic rounding, aligned activations and 1x16 weight blocks instead of NVIDIA recipe designNVFP4 linear layer · OLMo2-370M, 212B tokensvalidation perplexity: 26.20 -> 25.50Improved

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