Page 4 / 25
291 posts in total. Keep on posting.
Showing posts 37–48 of 291. Each entry opens locally on this site; legacy Hexo posts link back to their original article at the bottom for reference.
2026
- 中
DynaCalKV 阅读笔记：低秩 KV Cache 压缩里，固定分组到底有多不合理
低秩 KV cache 压缩里对注意力头做固定大小分组这件事,原来只是个没必要的假设——让相似度结构自己决定分组,确实能省参数,但前提是这个模型本来就有足够多可聚类的独立头。
- EN
Libra: Taming Attention Workload Skew in Long-Context LLM Training with Bounded Sequence Pool
Sequence packing balances memory, not attention FLOPs — Libra shows that the right fix is to keep the load-balancing pool a fixed size, tuned by the law of large numbers, rather than letting it grow with the cluster.
- 中
Libra 阅读笔记:用有界序列池驯服长上下文训练中的注意力负载偏斜
序列打包只能均衡内存,均衡不了注意力算力——Libra 告诉我们,真正该做的是把负载均衡池的规模固定住,用大数定律去算这个固定值该多大,而不是让它随集群一起膨胀。
- EN
LOCKS: Page-Local Compact Key Summaries for Efficient Long-Context Decoding
A page-by-page eigenbasis, not a shared low-rank sketch, is what lets a resident KV-cache summary track the exact top-k oracle down to the smallest attention budgets — here is why the scope of a summary matters more than its size.
- 中
LOCKS 阅读笔记：为什么 KV Cache 摘要要按「页」而不是按「序列」或「全局」来建
同样大小的摘要，只要拟合的范围（scope）从全局或整段序列缩小到单个 page，选中关键 token 的能力就有质的差别——这篇笔记梳理 LOCKS 论文里这条推理链的每一环。
- EN
Branching Policy Optimization: Turning Sandbox Snapshots Into a Variance-Reduction Primitive
BPO forks sibling rollouts from snapshotted mid-trajectory states instead of only the prompt, turning a sandbox's checkpoint-restore capability into a provably lower-variance advantage estimator than GRPO or RLOO.
- 中
Branching Policy Optimization：把沙盒快照变成一个方差削减原语
BPO 从沙盒中途的快照状态而不只是从 prompt 出发去分叉兄弟 rollout，把沙盒的 checkpoint-restore 能力变成了一个可证明比 GRPO、RLOO 更低方差的 advantage 估计器。
- EN
The Regression Tax: Why Adding Skills to an LLM Agent Also Breaks Tasks It Used to Solve
A paired-trajectory study of 5,832 agent runs showing that skill libraries buy their gains at a steep, usually invisible cost: 59% of the tasks they newly solve are offset by tasks they newly break, and the breakage traces to three concrete, nameable mechanisms.
- 中
Regression Tax（回归税）：给 LLM Agent 加 Skill 为什么既能帮忙又能拆台
一篇基于 5832 次配对运行的实证研究：Agent Skill 库带来的收益里，有 59% 会被它自己制造的新失败悄悄抵消，而这种破坏可以归因到三种可以在执行轨迹里指认出来的具体机制。
- EN
KV-Fold: Turning the KV Cache Into a Left Fold for Long-Context Inference
A training-free protocol that lets a frozen transformer carry its own KV cache across chunks like a left-fold accumulator, holding onto exact long-range facts where sliding-window streaming quietly forgets them.
- 中
KV-Fold：把 KV 缓存当作长上下文推理的左折叠累加器
一种无需训练的推理协议：让冻结的 transformer 像函数式编程里的左折叠一样，把 KV 缓存作为累加器在 chunk 间传递，在滑窗流式方法悄悄遗忘的地方保持精确的长距离事实检索。
- EN
When Does Adding More Agents Actually Help? An Information-Bottleneck Answer
This paper shows that multi-agent LLM systems only beat single-agent systems when their inter-worker relay messages compress away irrelevant context without discarding what the next worker actually needs, formalizing the trade-off as an information bottleneck controlled by a capability-dependent parameter.