# Learning What to Remember: Long-horizon Counterfactual Memory Optimization

arXiv: https://arxiv.org/abs/2609.37930 · HTML: https://mlsanity.com/paper/2609.37930/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 42 of its 48 references that are in the map.
- First posted on arXiv 29 Sep 2026.

## Builds on (most cited first, 30 of 42)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Let's Verify Step by Step](https://mlsanity.com/paper/2305.20050/) (2023)
- [A-MEM: Agentic Memory for LLM Agents](https://mlsanity.com/paper/2502.12110/) (2025)
- [Math-Shepherd: Verify and Reinforce LLMs Step-by-step without Human Annotations](https://mlsanity.com/paper/2312.08935/) (2023)
- [Mem0: Building Production-Ready AI Agents with Scalable Long-Term Memory](https://mlsanity.com/paper/2504.19413/) (2025)
- [Counterfactual Multi-Agent Policy Gradients](https://mlsanity.com/paper/1705.08926/) (2017)
- [Phi-4 Technical Report](https://mlsanity.com/paper/2412.08905/) (2024)
- [LightRAG: Simple and Fast Retrieval-Augmented Generation](https://mlsanity.com/paper/2410.05779/) (2024)
- [Rewarding Progress: Scaling Automated Process Verifiers for LLM Reasoning](https://mlsanity.com/paper/2410.08146/) (2024)
- [MemAgent: Reshaping Long-Context LLM with Multi-Conv RL-based Memory Agent](https://mlsanity.com/paper/2507.02259/) (2025)
- [DocRED: A Large-Scale Document-Level Relation Extraction Dataset](https://mlsanity.com/paper/1906.06127/) (2019)
- [MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon Agents](https://mlsanity.com/paper/2506.15841/) (2025)
- [Memory-R1: Enhancing Large Language Model Agents to Manage and Utilize Memories via Reinforcement Learning](https://mlsanity.com/paper/2508.19828/) (2025)
- [BABILong: Testing the Limits of LLMs with Long Context Reasoning-in-a-Haystack](https://mlsanity.com/paper/2406.10149/) (2024)
- [Recurrent Memory Transformer](https://mlsanity.com/paper/2207.06881/) (2022)
- [RUDDER: Return Decomposition for Delayed Rewards](https://mlsanity.com/paper/1806.07857/) (2018)
- [Document-Level Relation Extraction with Adaptive Thresholding and Localized Context Pooling](https://mlsanity.com/paper/2010.11304/) (2020)
- [SciREX: A Challenge Dataset for Document-Level Information Extraction](https://mlsanity.com/paper/2005.00512/) (2020)
- [Reasoning with Latent Structure Refinement for Document-Level Relation Extraction](https://mlsanity.com/paper/2005.06312/) (2020)
- [Hindsight Credit Assignment](https://mlsanity.com/paper/1912.02503/) (2019)
- [Agentic Memory: Learning Unified Long-Term and Short-Term Memory Management for Large Language Model Agents](https://mlsanity.com/paper/2601.01885/) (2026)
- [Connecting the Dots: Document-level Neural Relation Extraction with Edge-oriented Graphs](https://mlsanity.com/paper/1909.00228/) (2019)
- [Counterfactual Credit Assignment in Model-Free Reinforcement Learning](https://mlsanity.com/paper/2011.09464/) (2020)
- [CompAct: Compressing Retrieved Documents Actively for Question Answering](https://mlsanity.com/paper/2407.09014/) (2024)
- [Optimizing Agent Behavior over Long Time Scales by Transporting Value](https://mlsanity.com/paper/1810.06721/) (2018)
- [Revisiting DocRED -- Addressing the False Negative Problem in Relation Extraction](https://mlsanity.com/paper/2205.12696/) (2022)
- [PRCA: Fitting Black-Box Large Language Models for Retrieval Question Answering via Pluggable Reward-Driven Contextual Adapter](https://mlsanity.com/paper/2310.18347/) (2023)

## Abstract (arXiv)

Persistent textual memory allows language models to carry information across long interactions, but learning what to remember is fundamentally a credit-assignment problem. A memory rewrite may only become useful many steps later, while much of the observed utility may be inherited from information already stored before the rewrite. We introduce Memory Gain Policy Optimization (MGPO), which isolates the incremental value of each memory rewrite by crediting it for its marginal contribution to current and future downstream utility. This turns delayed memory utility into a direct learning signal for optimizing what information should persist. We study MGPO on document-level information extraction, where structured supervision makes the effects of individual memory updates directly measurable. MGPO improves extraction while reducing average memory length by nearly 80% relative to the initial memory policy before optimization. The learned memory policy also supports reuse and transfer across domains, downstream models without further training. These results show that effective memory learning depends not only on preserving useful information, but on identifying which memory updates create lasting incremental value.