paper-with-me

홈 › Papers

Why Gated DeltaNet Survives 4-Bit Quantization: NVFP4 W4A4 for the Recurrent Half of a Hybrid 27B LLM

2026-09-03 · Sergii Kozyrev, Davyd Maiboroda hf

Hybrid LLMs pair softmax attention with linear-attention layers such as Gated DeltaNet (GDN), whose recurrent state summarizes the context in fixed size. Early community 4-bit quantizations of Qwen3.8-27B (48 GDN layers, 16 attention layers) left the GDN block in 8- or 16-bit precision -- especially its decay and write-strength gates -- on the intuition that errors in a recurrence accumulate over long contexts. We test that intuition by building Minima: NVFP4 W4A4 on all 496 linear layers, GDN included. Across perplexity at 4K/32K, MMLU-Pro, GSM8K, AIME'25, GPQA-Diamond, LiveCodeBench, and RULER retrieval to 64K, Minima matches BF16 within seed noise (5-task average -0.52) while being the smallest (17.5 GiB) and fastest-prefill (+14-19%) recipe we compare, and its 32K perplexity gap shrinks with position. A four-part mechanism study explains why: (i) NVFP4's 16-element block scaling localizes the residual stream's extreme outliers, equalizing activation error across layer roles; (ii) the supposedly fragile gate projections are the least sensitive -- softplus/exponential and sigmoid parameterizations compress ~11% GEMM error to ~2% output error; (iii) the delta-rule recurrence holds injected noise at a flat plateau over 32K tokens and forgets a state impulse within hundreds of steps, because each write overwrites the state along the current key direction; (iv) the per-token quantization cost washes out with context instead of compounding. We also repair a global-scale mismatch that arises when per-module-calibrated NVFP4 checkpoints are served by kernels that fuse those modules into one GEMM, and show calibrated FP8 KV-cache scales are performance-free. The result: a practical recipe -- quantize everything, ship KV scales -- and a mechanistic account of why the recurrent half of a hybrid LLM is the easy half to quantize. Checkpoint: https://huggingface.co/minima-ai/mnma_qwen3.8_27b_nvfp4

📄 PDF Abstract BibTeX arXiv:2609.04098

Code (1)

🤗 minima-ai/mnma_qwen3.8_27b_nvfp4

Similar Papers 제목 키워드 기반

Provably Shorter Scratchpads in Hybrid DeltaNet-Attention Decoders

2026-05-15 · Tomasz Steifer arxiv

We investigate the expressive power of hybrid recurrent-attention decoders, a class of architectures used in recent open-source language models such as Qwen3-Next and its successors. These models combine Gated Attention …

Gated DeltaNet-2: Decoupling Erase and Write in Linear Attention

2026-05-21 · Ali Hatamizadeh, Yejin Choi, Jan Kautz arxiv

Linear attention replaces the unbounded cache of softmax attention with a fixed-size recurrent state, reducing sequence mixing to linear time and decoding to constant memory. The hard part is not just what to forget, but…

Linear Attention Architectures: Mechanisms, Trade-offs, and Cross-Layer Routing

2026-07-08 · Tommaso Cerruti, Tim Rieder, George Rowlands, Lingfeng Jin 외 arxiv

Self-attention lets each token retrieve information from the full context, but its quadratic cost in sequence length limits training and inference at long context. This paper presents a comparative study of softmax atten…

Full-Stack FP4: Stable LLM Pretraining with Quantized Projections, Optimizers, and Attention

2026-07-05 · Siyu Ding, Mingchuan Ma, Jiabo Tong, Xingrun Xing 외 arxiv

Recent NVFP4 pretraining methods mainly target transformer linear layers, leaving optimizer states, optimizer arithmetic and attention underexplored in 4-bit pipelines. This critical gap blocks stable full-stack 4-bit pr…

Four Over Six: More Accurate NVFP4 Quantization with Adaptive Block Scaling

2025-12-01 · Jack Cook, Junxian Guo, Guangxuan Xiao, Yujun Lin 외 arxiv

As large language models have grown larger, interest has grown in low-precision numerical formats such as NVFP4 as a way to improve speed and reduce memory usage. However, quantizing models to NVFP4 remains challenging a…