Page 13 / 25
293 posts in total. Keep on posting.
Showing posts 145–156 of 293. Each entry opens locally on this site; legacy Hexo posts link back to their original article at the bottom for reference.
2026
- 中
KVQuant：面向千万级上下文的 KV 缓存量化技术——阅读笔记
KVQuant（NeurIPS 2024，UC Berkeley）通过逐通道键量化、预-RoPE 量化、Fisher 加权非均匀量化数据类型（nuqX）和逐向量稠密-稀疏残差，实现 sub-4-bit KV 缓存量化，将上下文扩展至千万级别，3-bit 困惑度退化仅 0.07。
- EN
REINFORCE++: Stabilizing Critic-Free Policy Optimization with Global Advantage Normalization
REINFORCE++ proves that GRPO's per-prompt advantage normalization is a biased estimator, then fixes it with a single global batch normalization step — achieving state-of-the-art results across general RLHF, complex reasoning, and long-horizon agentic tasks, all without a critic network.
- 中
REINFORCE++：用全局优势归一化稳定免批评家策略优化
REINFORCE++ 从数学上证明了 GRPO 的逐 prompt 局部归一化是一个有偏估计量，并用全局批次归一化替换它——在通用 RLHF、复杂推理和长时序 agent 任务上全面超越 GRPO 和 PPO，同时无需任何批评家网络。
- EN
AutoSci: A Memory-Centric Agentic System for the Full Scientific Research Lifecycle
AutoSci is a memory-centric agentic system that automates the full scientific research lifecycle — reading, ideation, experimentation, writing, and rebuttal — through four integrated modules (SciMem, SciFlow, SciDAG, SciEvolve). I walk through the architecture from scratch and critically assess its evaluation methodology and narrow domain coverage.
- 中
AutoSci：以记忆为中心的全科研生命周期自主智能体系统
AutoSci 是北大团队提出的「永久性科研环境」，用以记忆为中心的多智能体把读文献、提想法、做实验、写论文、回审稿人串成一个能自我进化的闭环。本文从零梳理它的四大模块（SciMem / SciFlow / SciDAG / SciEvolve）与两个端到端案例，并批判性分析其评测方法与适用边界上的局限。
- EN
Group Sequence Policy Optimization: A Sequence-Level RL Algorithm for Training Large Language Models
GSPO replaces GRPO's token-level importance ratios with a single sequence-level ratio, yielding more stable and efficient RL for LLMs — and crucially fixing the training collapse that plagues RL on large Mixture-of-Experts models. A from-scratch walkthrough of the math and algorithm, plus a critical look at what the paper leaves untested.
- 中
Group Sequence Policy Optimization：序列级重要性采样修正 GRPO 的 RL 训练方法
GSPO 把 GRPO 的 token 级重要性比率换成单一的序列级比率，让 LLM 的强化学习训练更稳、更省，并解决了大型 MoE 模型上 RL 训练崩溃的难题。本文从零讲清它的数学动机与算法细节，并批判性地分析了论文尚未验证的部分。
- EN
VeriCache: Turning Lossy KV Cache into Lossless LLM Inference
VeriCache reframes KV cache compression as a speculative decoding problem: draft with lossy compressed KV, verify with full KV cache, and accept matched tokens — achieving up to 4× throughput while guaranteeing bit-identical outputs.
- 中
VeriCache：让有损 KV 缓存实现无损 LLM 推理
VeriCache 把 KV 缓存压缩重新包装成投机解码问题：用压缩 KV 生成草稿 token，再用完整 KV 校验，保证输出与全量 KV 推理完全一致，同时实现最高 4× 吞吐量提升。
- EN
IO-SVD: Input-Output Whitened SVD for Adaptive-Rank LLM Compression
How double-sided KL-aware whitening, adaptive heterogeneous rank allocation, and loss-aware remapping combine to push SVD-based LLM compression to a new state of the art — with 4.34× decode throughput and minimal quality loss even at 60% parameter removal.
- 中
IO-SVD：基于输入输出双侧白化的自适应秩LLM压缩方法
KL散度感知的双侧白化 + 贪婪异构秩分配 + 损失感知量化重映射，三招组合将SVD压缩推到新的SOTA——在LLaMA-7B 80%保留率下PPL降至5.59，同时带来4.34倍解码吞吐提升。
- EN
Mooncake: A KVCache-centric Disaggregated Architecture for LLM Serving
How Moonshot AI's Kimi serving platform redesigned LLM infrastructure around KV cache disaggregation—achieving 525% throughput gains for long-context workloads while maintaining strict TTFT and TBT SLO compliance.