# TRACER: Per-Tool Context Retention for LLM Agents via Consequence-Attributed Reinforcement Learning

arXiv: https://arxiv.org/abs/2608.29363 · HTML: https://mlsanity.com/paper/2608.29363/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning.
- Builds on 19 of its 30 references that are in the map.
- First posted on arXiv 29 Aug 2026.

## Placed in

- LLM reasoning / [Reasoning for agents, planning and code](https://mlsanity.com/lr/reasoning-for-agents-planning-and-code/)
- LLM agents and tool use / [Tool learning and function calling](https://mlsanity.com/agents/tool-learning-and-function-calling/)

## Builds on (most cited first, 19 of 19)

- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [MemGPT: Towards LLMs as Operating Systems](https://mlsanity.com/paper/2310.08560/) (2023)
- [Counterfactual Multi-Agent Policy Gradients](https://mlsanity.com/paper/1705.08926/) (2017)
- [LongLLMLingua: Accelerating and Enhancing LLMs in Long Context Scenarios via Prompt Compression](https://mlsanity.com/paper/2310.06839/) (2023)
- [LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compression](https://mlsanity.com/paper/2403.12968/) (2024)
- [RUDDER: Return Decomposition for Delayed Rewards](https://mlsanity.com/paper/1806.07857/) (2018)
- [ACON: Optimizing Context Compression for Long-horizon LLM Agents](https://mlsanity.com/paper/2510.00615/) (2025)
- [Scaling LLM Multi-turn RL with End-to-end Summarization-based Context Management](https://mlsanity.com/paper/2510.06727/) (2025)
- [Sculptor: Empowering LLMs with Cognitive Agency via Active Context Management](https://mlsanity.com/paper/2508.04664/) (2025)
- [Learning Agent-Compatible Context Management for Long-Horizon Tasks](https://mlsanity.com/paper/2605.30785/) (2026)
- [LOCA-bench: Benchmarking Language Agents Under Controllable and Extreme Context Growth](https://mlsanity.com/paper/2602.07962/) (2026)
- [CM2: Reinforcement Learning with Checklist Rewards for Multi-Turn and Multi-Step Agentic Tool Use](https://mlsanity.com/paper/2602.12268/) (2026)
- [Understanding Individual Agent Importance in Multi-Agent System via Counterfactual Reasoning](https://mlsanity.com/paper/2412.15619/) (2024)
- [Verifiable Process Rewards for Agentic Reasoning](https://mlsanity.com/paper/2605.10325/) (2026)
- [Enhancing Agentic RL with Progressive Reward Shaping and Value-based Sampling Policy Optimization](https://mlsanity.com/paper/2512.07478/) (2025)
- [When Denser Credit Is Not Enough: Evidence-Calibrated Policy Optimization for Long-Horizon LLM Agent Training](https://mlsanity.com/paper/2606.05885/) (2026)
- [Learning When Not to Act: Mitigating Tool Abuse in Agentic Reinforcement Learning](https://mlsanity.com/paper/2606.02132/) (2026)
- [AGORA: Adapter-Grounded Observation-Action Retention for Inference-Free Prompt Compression in LLM Agents](https://mlsanity.com/paper/2605.26596/) (2026)
- [Tool-Aware Optimization with Entropy Guidance for Efficient Agentic Reinforcement Learning](https://mlsanity.com/paper/2606.03762/) (2026)

## Built on by (most cited first, 1 of 1)

- [AgentBetta: Verification-Driven Adaptive Configuration of an AI Nano-Agent through Selective Expansion and Verified Contraction](https://mlsanity.com/paper/2609.23512/) (2026)

## Abstract (arXiv)

Enterprise data agents answer business queries by chaining many tool calls over multiple reasoning steps, routinely accumulating hundreds of thousands of context tokens per session. Existing compression strategies typically allocate retention budgets without accounting for the downstream consequences of removing individual tool outputs. Aggressive compression may therefore trigger costly tool re-invocations that offset the initial savings. We call this the compression--consequence gap. To close it, we propose TRACER, which formulates compression as a sequential per-tool decision problem. A lightweight REINFORCE policy assigns query-conditioned retention ratios using only information available at each compression event. Its consequence-aware objective jointly accounts for task success, total token consumption, and post-compression tool re-invocations. To improve credit assignment, TRACER uses a learned outcome model to compare the predicted consequences of the selected retention ratio with those of fully retaining each tool output. On held-out production queries across three compressor backends, TRACER reduces total token consumption by 29--46% relative to keeping all context while maintaining comparable or higher task success. Compared with a tool-type-conditional static policy, TRACER provides an additional 15--18% of token savings. Interventional rollouts show that the learned per-tool credit scores correlate with measured single-tool consequences. The learned policy also yields positive savings when transferred across agent backbones and compressor architectures, and reduces token consumption by 18--25% on five held-out LOCA-bench environments. These results demonstrate the value of consequence-aware, per-tool context retention for improving the efficiency of long-horizon language agents.