# AWM: Answerable Working Memory for Long-Document VQA Agents

arXiv: https://arxiv.org/abs/2608.25618 · HTML: https://mlsanity.com/paper/2608.25618/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Multimodal reasoning, one of 17 directions in LLM reasoning.
- Builds on 36 of its 44 references that are in the map.
- First posted on arXiv 26 Aug 2026.

## Placed in

- LLM reasoning / [Multimodal reasoning](https://mlsanity.com/lr/multimodal-reasoning/)
- LLM agents and tool use / [Agent memory systems](https://mlsanity.com/agents/agent-memory-systems/)

## Builds on (most cited first, 30 of 36)

- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection](https://mlsanity.com/paper/2310.11511/) (2023)
- [ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks](https://mlsanity.com/paper/2303.15056/) (2023)
- [Enabling Large Language Models to Generate Text with Citations](https://mlsanity.com/paper/2305.14627/) (2023)
- [Evaluating Verifiability in Generative Search Engines](https://mlsanity.com/paper/2304.09848/) (2023)
- [VisRAG: Vision-based Retrieval-augmented Generation on Multi-modality Documents](https://mlsanity.com/paper/2410.10594/) (2024)
- [Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Models](https://mlsanity.com/paper/2401.06102/) (2024)
- [MMLongBench-Doc: Benchmarking Long-context Document Understanding with Visualizations](https://mlsanity.com/paper/2407.01523/) (2024)
- [M3DocRAG: Multi-modal Retrieval is What You Need for Multi-page Multi-document Understanding](https://mlsanity.com/paper/2411.04952/) (2024)
- [LongDocURL: a Comprehensive Multimodal Long Document Benchmark Integrating Understanding, Reasoning, and Locating](https://mlsanity.com/paper/2412.18424/) (2024)
- [VRAG-RL: Empower Vision-Perception-Based RAG for Visually Rich Information Understanding via Iterative Reasoning with Reinforcement Learning](https://mlsanity.com/paper/2505.22019/) (2025)
- [Perception-R1: Advancing Multimodal Reasoning Capabilities of MLLMs via Visual Perception Reward](https://mlsanity.com/paper/2506.07218/) (2025)
- [Unlocking Exploration in RLVR: Uncertainty-aware Advantage Shaping for Deeper Reasoning](https://mlsanity.com/paper/2510.10649/) (2025)
- [Visual Instance-aware Prompt Tuning](https://mlsanity.com/paper/2507.07796/) (2025)
- [MemSearcher: Training LLMs to Reason, Search and Manage Memory via End-to-End Reinforcement Learning](https://mlsanity.com/paper/2511.02805/) (2025)
- [Look Inward to Explore Outward: Learning Temperature Policy from LLM Internal States via Hierarchical RL](https://mlsanity.com/paper/2602.13035/) (2026)
- [SV-RAG: LoRA-Contextualizing Adaptation of MLLMs for Long Document Understanding](https://mlsanity.com/paper/2411.01106/) (2024)
- [Docopilot: Improving Multimodal Models for Document-Level Understanding](https://mlsanity.com/paper/2507.14675/) (2025)
- [MoLoRAG: Bootstrapping Document Understanding via Multi-modal Logic-aware Retrieval](https://mlsanity.com/paper/2509.07666/) (2025)
- [ArgRAG: Explainable Retrieval Augmented Generation using Quantitative Bipolar Argumentation](https://mlsanity.com/paper/2508.20131/) (2025)
- [When TableQA Meets Noise: A Dual Denoising Framework for Complex Questions and Large-scale Tables](https://mlsanity.com/paper/2509.17680/) (2025)
- [Predictive Multiplicity of Knowledge Graph Embeddings in Link Prediction](https://mlsanity.com/paper/2408.08226/) (2024)
- [What Breaks Knowledge Graph based RAG? Benchmarking and Empirical Insights into Reasoning under Incomplete Knowledge](https://mlsanity.com/paper/2508.08344/) (2025)
- [Auditing Data Membership in Reinforcement Learning With Verifiable Rewards](https://mlsanity.com/paper/2511.14045/) (2025)
- [Rethinking Table Pruning in TableQA: From Sequential Revisions to Gold Trajectory-Supervised Parallel Search](https://mlsanity.com/paper/2601.03851/) (2026)
- [Evaluating Knowledge Graph Based Retrieval Augmented Generation Methods under Knowledge Incompleteness](https://mlsanity.com/paper/2504.05163/) (2025)
- [Enrich-on-Graph: Query-Graph Alignment for Complex Reasoning with LLM Enriching](https://mlsanity.com/paper/2509.20810/) (2025)
- [Supposedly Equivalent Facts That Aren't? Entity Frequency in Pre-training Induces Asymmetry in LLMs](https://mlsanity.com/paper/2503.22362/) (2025)
- [One Refiner to Unlock Them All: Inference-Time Reasoning Elicitation via Reinforcement Query Refinement](https://mlsanity.com/paper/2604.25444/) (2026)

## Built on by (most cited first, 2 of 2)

- [MemoryWalker: Stop Training Agents on Contexts They Never Saw](https://mlsanity.com/paper/2609.00865/) (2026)
- [Uncertainty in Representation Learning on Knowledge Graphs](https://mlsanity.com/paper/2610.06974/) (2026)

## Abstract (arXiv)

Long-document visual question answering increasingly relies on VLM agents that retrieve candidate pages, inspect page images, write findings to working memory, and synthesize answers. Working memory should carry answer-supporting evidence across page inspections for later grounded answering, yet existing evaluation mainly checks final-answer correctness and evidence-page access. This creates a memory-quality blind spot: an agent may reach the right page and answer correctly while leaving behind memory too generic or incomplete to support answering once page context is removed. We introduce memory-only answerability, a diagnostic that asks whether a reader can answer from the question and terminal working memory alone. Building on this diagnostic, Answerable Working Memory (AWM) treats terminal working memory as an answerable evidence artifact, and AWM-GRPO incorporates this signal into the GRPO reward while preserving final-answer priority. Under GRPO, this reward assigns higher advantages to answer-correct trajectories whose terminal working memory remains answerable. On textscMMLongBench-Doc, even when gold evidence pages are provided, 42.5% of correct answers still cannot be answered from terminal working memory alone. AWM-GRPO improves final-answer accuracy over the RAG baseline by 8.1 and 11.9 points on textscMMLongBench-Doc and textscLongDocURL and reduces the memory-missing-correct rate by 2.7 points over answer-only GRPO.