Tag
该论文研究在单一人类注意力有限、多个 AI 代理并行执行任务时，代理应在何时自主搜索信息、何时向人类请求澄清，并提出 Pairwise Coordination Policy (PCP)，在合成基准与 LLM 案例研究中显著降低成本。
A new arXiv paper introduces Functional Rank Allocation (FRA), an unlabeled method for compressing large fleets of LoRA adapters via functional rank-distortion frontiers and fleet-level minimax allocation under a shared rank budget, improving worst-case and tail utility over uniform spectral compression across 78 adapters from three public fleets.
TAO 提出一个工具感知的联合调度框架，用于基于 LLM 的智能体服务，通过动态规划在容量约束下选择和放置会话，相比最优基线最高降低 26.8% 延迟、36.3% TTFT 并提升 29.5% 解码吞吐。
OOPSLA 2026 paper introducing statewalk DP, a new extraction algorithm for effectful e-graphs that enforces effect ordering without external ILP solvers, achieving order-of-magnitude speedups in the eggcc prototype compiler.
提出 VISTA 这一免训练方法，通过在少量全引导 warmup 生成中收集价值统计并用动态规划优化，在固定预算下自动为离散扩散的 SMC 奖励引导选择引导时间步，在文本、生物序列和图像生成任务上优于手工稀疏调度，实现最高 5.35× 加速。
The article introduces dp_knapsack_sliding_hirschberg, a new activation memory budget solver for PyTorch that reduces peak RAM usage during memory planning by up to 20x compared to the default dp_knapsack solver, explaining how 0/1 knapsack dynamic programming is used to choose which operations to save or recompute.
A technical deep-dive blog series exploring join ordering in query optimization, focusing on the German school of dynamic programming approaches (DPsize, DPsub, DPccp, DPhyp) and related work on adaptive optimization and join search space enumeration.
The paper introduces Neural Spectral Capacity (NSC), a training-free metric based on the singular-value spectrum to evaluate and optimize neural network architectures, with a dynamic programming method for optimal design under constraints.
This paper proposes Fixed-Range Chunking (FRC) for document-level machine translation to ensure length consistency, reducing train-test mismatch and improving translation quality in large language models.
This paper introduces a joint feature and sample reduction framework to scale optimal classification tree learning, achieving significant speedups while maintaining predictive performance.
This paper investigates why standard neural architectures fail to generalize to longer inputs in dynamic programming, using geometric analysis with tropical semiring theory to reveal structural limitations in compositions.
A deep-dive educational article exploring dynamic programming as a unifying principle behind shortest-path algorithms, neural network training, and context-free grammar parsing, tying together automata, optimal control, and linear programming.
This paper proposes a method using reservoir computing to recycle computational processes of dynamic programming for combinatorial optimization problems, achieving improved approximation accuracy and reduced computation time on traveling salesman and subset sum problems.
The post highlights a Dynamic Programming formulation from the GLM-5.2 paper's appendix that resembles a LeetCode problem, tying into the debate about LeetCode's relevance sparked by Alisa Liu's hiring at OpenAI.
This paper introduces Bellman-Taylor Score Decoding, a method to handle state-dependent feasible action sets in Markov decision processes, addressing a key challenge in applying deep reinforcement learning to operations research problems.
A novel method for multilingual word-level forced alignment combines self-supervised representations from MMS and a phoneme boundary detector with a learned dynamic programming decoder, outperforming existing aligners on English and unseen languages without further training.
The article explores the optimal ApplicativeDo scheduling problem in GHC, which is disabled by default due to slowness, and draws an analogy to the dynamic programming algorithm used for RNA folding to improve compiler performance.
A Twitter thread uses a Google coding interview example to highlight the high compensation at Google and then pivots to how AI agents like Claude Skills can replace entire dev teams, changing the job market.
This paper presents a hybrid approach combining dynamic programming and constraint programming to solve the Partial Shop Scheduling Problem, demonstrating the viability of integrating both paradigms despite not outperforming pure CP solvers.
This paper proposes a unified knowledge-embedded reinforcement learning framework for generalized capacitated vehicle routing problems, combining route-first cluster-second heuristics with dynamic programming to achieve superior solution quality and strong generalization across diverse variants.