Tag
介绍一种基于768维嵌入空间上的吸收马尔可夫链（Absorbing Markov Chain）的语义分流引擎，通过在60节点领域流形上做随机游走来替代朴素余弦相似度路由，实现<2ms的零token意图分类，从而在唤醒LLM之前完成路由。文中给出JevBench基准测试结果：四层级联系统（Gevva0）在240个条目上达到100%召回、89.17%精度，平均p50延迟46ms，并配备否定/语法、流形邻近OOD和熵与边界三重防护网。
作者将 Gevva0 升级为级联决策引擎（静态规则 → 吸收马尔可夫链 → 快速路径 logits → 有界 CoT），并诚实地发布了对先前"100% GTO"说法的更正：修复提示泄漏 bug 后，零样本 GTO 一致率实为约 15%-29%，但延迟仍保持约 2ms。
This paper introduces TAM-Chain, a multi-scale thyroid cytology classification method that employs absorbing Markov chains and Shannon entropy for uncertainty quantification, aiming to suppress false negatives and adapt to domain shifts.
This paper studies transformers trained on Markov chains and identifies that induction heads implement soft context-matching and Dirichlet-style smoothing, showing that transformers regularize in-context estimation rather than simply counting n-grams.
Mathematicians have extended the classic 1992 proof about card shuffling to less precise shuffles, showing that a 'cutoff phenomenon' still occurs even with uneven deck splits.
A tweet shares a lecture at MIT by David Shirokoff covering the fundamentals of Markov Chains, including transition probabilities, Markov matrices, eigenvalues, and long-term steady state.
This paper analyzes the global distributional behavior induced by iterative masked-token resampling in masked language models using Glauber dynamics. It introduces a rectangle test for incompatibility, establishes mixing time bounds, and empirically demonstrates phase transitions and metastable semantic basins.
Jane Street offers $750k/year for quants who can apply Stochastic Processes and Markov Chains in trading, and a free MIT lecture covers similar material.
Researchers use four-state Markov chains to model vowel/consonant patterns in Pushkin’s Evgenij Onegin and its Italian translation, revealing structural asymmetries and narrative-linked phonological cues.