TVCACHE: A Stateful Tool-Value Cache for Post-Training LLM Agents

• URL: http://arxiv.org/abs/2602.10986v1
• Date: Wed, 11 Feb 2026 16:13:44 GMT
• Title: TVCACHE: A Stateful Tool-Value Cache for Post-Training LLM Agents
• Abstract summary: TVCACHE is a stateful tool-value cache for LLM agent post-training.<n>Hit rates of up to 70% and reduces median tool call execution time by up to 6.9X.
• Score: 2.1447191409339244
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: In RL post-training of LLM agents, calls to external tools take several seconds or even minutes, leaving allocated GPUs idle and inflating post-training time and cost. While many tool invocations repeat across parallel rollouts and could in principle be cached, naively caching their outputs for reuse is incorrect since tool outputs depend on the environment state induced by prior agent interactions. We present TVCACHE, a stateful tool-value cache for LLM agent post-training. TVCACHE maintains a tree of observed tool-call sequences and performs longest-prefix matching for cache lookups: a hit occurs only when the agent's full tool history matches a previously executed sequence, guaranteeing identical environment state. On three diverse workloads-terminal-based tasks, SQL generation, and video understanding. TVCACHE achieves cache hit rates of up to 70% and reduces median tool call execution time by up to 6.9X, with no degradation in post-training reward accumulation.

Related papers

• Joint Optimization of Tool Creation and Use for Large Language Model Agents [69.07066297088187]
  SMITH (-grounded Multi-task Iterative Tool Honing) trains tool creation and tool use inside a single policy.<n>A 4B Qwen3 trained with SMITH on 13 procedural reasoning tasks with exact verifiers reaches 79.8 macro-average accuracy on held-out tasks.<n>It also reaches 40.4 on TabMWP-Hard and 42.6 on out-of-domain GQA (+7.6 over the best same-backbone inference-time baseline)
  arXiv  Detail & Related papers  (2026-08-25T13:59:34Z)
• MCP-Universe RL: A Framework for Training MCP Tool-Use Agents via Reinforcement Learning [54.082979654934]
  Reinforcement learning (RL) has become an effective way to improve the tool-use ability of large language models (LLMs)<n>Most existing RL frameworks stop at the policy update.<n>We present MCP-Universe RL (MCP-U RL), an open-source framework that takes over both.
  arXiv  Detail & Related papers  (2026-08-23T01:49:04Z)
• Kalypso: Relational LLM Serving [6.264258565843104]
  Existing semantic query processing systems invoke request-centric LLM serving systems that are unaware of the query plan.<n>We present Kalypso, a relational LLM serving system that exposes an API for semantic query plans and executes them using an adaptive, memory-aware scheduling algorithm.<n>Our evaluation shows that Kalypso improves query completion time over baselines using request-centric LLM serving, with speedups up to 4.57x across diverse workloads.
  arXiv  Detail & Related papers  (2026-07-26T19:29:31Z)
• TraceLab: Characterizing Coding Agent Workloads for LLM Serving [16.402747744601612]
  We release a trace of roughly 4,300 coding-agent sessions, containing about 350,000 LLM steps and 430,000 tool calls.<n>Our analysis shows that coding-agent workloads feature long autonomous loops, long contexts with short outputs, diverse and heavily-tailed tool calls, and high but imperfect prefix cache hit rates.<n>These findings point to concrete opportunities for optimizing serving, including lower-overhead tool calling, append-length-aware prefill, semantic-aware tool-latency prediction, and improved KV-cache management around human-paced gaps.
  arXiv  Detail & Related papers  (2026-06-29T16:59:05Z)
• Multi-Agent Computer Use [72.79887808312706]
  We argue that we should move towards evaluating and building multi-agent computer use (MACU) systems.<n>We propose a general multi-agent setup in which a manager model decomposes computer use tasks as a directed acyclic graph (DAG)<n>At each iteration, the manager dispatches parallel CUA subagents to carry out nodes on the ready frontier of the DAG.
  arXiv  Detail & Related papers  (2026-06-01T01:29:36Z)
• Evaluating Temporal Semantic Caching and Workflow Optimization in Agentic Plan-Execute Pipelines [3.4989760172592264]
  Industrial asset operations are latency-sensitive because a single user query may require coordination over sensor data, work orders, failure modes, forecasting tools, and domain-specific agents.<n>We evaluate this problem on AssetOpsBench (AOB), an industrial agent benchmark whose plan-execute pipeline exposes repeated overhead from tool discovery, LLM planning, MCP tool execution, and final summarization.<n>We propose two complementary optimization layers for AOB plan-execute pipelines: a temporal semantic cache and a set of MCP workflow optimizations combining disk-backed tool-discovery caching and dependency-aware parallel step execution.
  arXiv  Detail & Related papers  (2026-05-20T02:30:07Z)
• Learning to Share: Selective Memory for Efficient Parallel Agentic Systems [49.78267008828593]
  Agentic systems solve complex tasks by coordinating multiple agents that iteratively reason, invoke tools, and exchange intermediate results.<n>Recent approaches deploy multiple agent teams running in parallel to explore diverse reasoning trajectories.<n>We propose Learning to Share (LTS), a learned shared-memory mechanism for parallel agentic frameworks.
  arXiv  Detail & Related papers  (2026-02-05T18:20:21Z)
• ToolCaching: Towards Efficient Caching for LLM Tool-calling [13.738787213936225]
  Caching is a classic solution to the problem of redundant or repeated tool-calling requests.<n>We propose ToolCaching, an efficient feature-driven and adaptive caching framework.<n>ToolCaching achieves up to 11% higher cache hit ratios and 34% lower latency compared to standard policies.
  arXiv  Detail & Related papers  (2026-01-20T09:25:59Z)
• Don't Break the Cache: An Evaluation of Prompt Caching for Long-Horizon Agentic Tasks [1.2292307778008844]
  We present a comprehensive evaluation of prompt caching across three major Large Language Model (LLM) providers.<n>Our results demonstrate that prompt caching reduces API costs by 45-80% and improves time to first token by 13-31% across providers.
  arXiv  Detail & Related papers  (2026-01-09T18:41:57Z)
• Quantifying Memory Use in Reinforcement Learning with Temporal Range [51.98491034847041]
  Temporal Range is a model-agnostic metric that treats first-order sensitivities of multiple vector outputs across a temporal window to the input sequence as a temporal influence profile.<n>We also report Temporal Range for a compact Long Expressive Memory (LEM) policy trained on the task, using it as a proxy readout of task-level memory.
  arXiv  Detail & Related papers  (2025-12-05T22:58:09Z)
• Continuum: Efficient and Robust Multi-Turn LLM Agent Scheduling with KV Cache Time-to-Live [30.099614426825834]
  Continuum is a serving system to optimize job completion time for multi-turn agent workloads.<n>By predicting tool call durations in agentic, Continuum selectively pins the KV cache in GPU memory with a time-to-live value based on total turn number.<n>Our evaluation on real-world agentic workloads with Llama-3.1 8B/70B models shows that Continuum significantly improves the average job completion times.
  arXiv  Detail & Related papers  (2025-11-04T03:43:05Z)
• Semantic-Aware Scheduling for GPU Clusters with Large Language Models [60.14838697778884]
  We propose SchedMate, a framework that bridges the semantic gap between schedulers and jobs they manage.<n>SchedMate extracts deep insights from overlooked, unstructured data sources: source code, runtime logs, and historical jobs.<n>We show SchedMate reduces average job completion times by up to 1.91x, substantially enhancing the scheduling performance.
  arXiv  Detail & Related papers  (2025-10-02T02:01:02Z)
• LaCache: Ladder-Shaped KV Caching for Efficient Long-Context Modeling of Large Language Models [52.56008278458534]
  LaCache is a training-free method for efficient and accurate generative inference of Large Language Models.<n>LaCache enables LLMs to address both of the critical challenges in long-range modeling: robust long-range capabilities and continuous generation without running out-of-memory.
  arXiv  Detail & Related papers  (2025-07-14T19:09:57Z)
• RT-Cache: Training-Free Retrieval for Real-Time Manipulation [13.338672711391999]
  RT-Cache is a training-free retrieval-as-control pipeline for robots.<n>It embeds the current frame to retrieve and replay multi-step snippets, replacing per-step model calls.<n>A hierarchical search keeps lookups sub-second at million scale, shifting cost from compute to storage.
  arXiv  Detail & Related papers  (2025-05-14T00:41:44Z)
• Trajectory Balance with Asynchrony: Decoupling Exploration and Learning for Fast, Scalable LLM Post-Training [71.16258800411696]
  Reinforcement learning (RL) is a critical component of large language model (LLM) post-training.<n>Existing on-policy algorithms used for post-training are inherently incompatible with the use of experience replay buffers.<n>We propose efficiently obtaining this benefit of replay buffers via Trajectory Balance with Asynchrony (TBA)
  arXiv  Detail & Related papers  (2025-03-24T17:51:39Z)
• Leveraging Approximate Caching for Faster Retrieval-Augmented Generation [6.674782158041247]
  We introduce Proximity, an approximate key-value cache that optimize the RAG workflow by leveraging similarities in user queries.<n>Instead of treating each query independently, Proximity reuses previously retrieved documents when similar queries appear.<n>Our experiments demonstrate that Proximity with our LSH scheme and a realistically-skewed MedRAG workload reduces database calls by 77.2% while maintaining database recall and test accuracy.
  arXiv  Detail & Related papers  (2025-03-07T15:54:04Z)
• NumS: Scalable Array Programming for the Cloud [82.827921577004]
  We present NumS, an array programming library which optimize NumPy-like expressions on task-based distributed systems. This is achieved through a novel scheduler called Load Simulated Hierarchical Scheduling (LSHS) We show that LSHS enhances performance on Ray by decreasing network load by a factor of 2x, requiring 4x less memory, and reducing execution time by 10x on the logistic regression problem.
  arXiv  Detail & Related papers  (2022-06-28T20:13:40Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.