# MemToC: Benchmarking Memory-Tool Conflict Resolution in Large Language Models

arXiv: https://arxiv.org/abs/2608.26295 · HTML: https://mlsanity.com/paper/2608.26295/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Agent memory systems, one of 16 directions in LLM agents and tool use.
- Builds on 22 of its 34 references that are in the map.
- First posted on arXiv 26 Aug 2026.

## Placed in

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

## Builds on (most cited first, 22 of 22)

- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2](https://mlsanity.com/paper/2408.05147/) (2024)
- [Entity-Based Knowledge Conflicts in Question Answering](https://mlsanity.com/paper/2109.05052/) (2021)
- [Adaptive Chameleon or Stubborn Sloth: Revealing the Behavior of Large Language Models in Knowledge Conflicts](https://mlsanity.com/paper/2305.13300/) (2023)
- [AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders](https://mlsanity.com/paper/2501.17148/) (2025)
- [DisentQA: Disentangling Parametric and Contextual Knowledge with Counterfactual Question Answering](https://mlsanity.com/paper/2211.05655/) (2022)
- [FaithEval: Can Your Language Model Stay Faithful to Context, Even If "The Moon is Made of Marshmallows"](https://mlsanity.com/paper/2410.03727/) (2024)
- [Steering Knowledge Selection Behaviours in LLMs via SAE-Based Representation Engineering](https://mlsanity.com/paper/2410.15999/) (2024)
- [Astute RAG: Overcoming Imperfect Retrieval Augmentation and Knowledge Conflicts for Large Language Models](https://mlsanity.com/paper/2410.07176/) (2024)
- [Context-DPO: Aligning Language Models for Context-Faithfulness](https://mlsanity.com/paper/2412.15280/) (2024)
- [ToolHop: A Query-Driven Benchmark for Evaluating Large Language Models in Multi-Hop Tool Use](https://mlsanity.com/paper/2501.02506/) (2025)
- [KnowPO: Knowledge-aware Preference Optimization for Controllable Knowledge Selection in Retrieval-Augmented Language Models](https://mlsanity.com/paper/2408.03297/) (2024)
- [When Tools Fail: Benchmarking Dynamic Replanning and Anomaly Recovery in LLM Agents](https://mlsanity.com/paper/2606.05806/) (2026)
- [Beyond Function Calling: Benchmarking Tool-Using Agents under Tool-Environment Unreliability](https://mlsanity.com/paper/2606.25819/) (2026)
- [The Tool-Overuse Illusion: Why Does LLM Prefer External Tools over Internal Knowledge?](https://mlsanity.com/paper/2604.19749/) (2026)
- [When the Tool Decides: LLM Agents Defer Blindly to Graph Neural Network Tools, and Stronger Backbones Defer More](https://mlsanity.com/paper/2606.14476/) (2026)
- [SHIFT: Gate-Modulated Activation Steering for Knowledge Conflict Mitigation in Retrieval-Augmented Generation](https://mlsanity.com/paper/2606.27786/) (2026)
- [Trust No Tool: Evaluating and Defending LLM Agents under Untrusted Tool Feedback](https://mlsanity.com/paper/2605.17453/) (2026)
- [Controlling Tool Use with Heading-Specific Activation Steering](https://mlsanity.com/paper/2607.05790/) (2026)
- [Diagnosing LLM Arbitration Behavior over Pre-evidence Epistemic States in RAG-based Fact-Checking](https://mlsanity.com/paper/2606.01120/) (2026)
- [From Context-Aware to Conflict-Aware: Generalizing Contrastive Decoding for Knowledge Conflict in LLMs](https://mlsanity.com/paper/2606.10298/) (2026)
- [Regime-Aware Peer Specialization for Robust RAG under Heterogeneous Knowledge Conflicts](https://mlsanity.com/paper/2606.30518/) (2026)

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

- [When LLMs Sit Above Diagnostic Tools: Unrealized Complementarity in Industrial Fault Diagnosis](https://mlsanity.com/paper/2610.05031/) (2026)

## Abstract (arXiv)

Tool-augmented LLMs must arbitrate between two fallible sources when a tool return conflicts with their parametric memory, yet existing evaluations measure source preference without establishing source correctness. We introduce MemToC, a controlled benchmark for post-tool-return arbitration with executable tools. MemToC comprises 6,504 evaluation episodes constructed from 542 quality-controlled factual questions, independently elicited model-specific closed-book answers, and controlled tool returns of known correctness. These components instantiate four source-correctness cases; tool-error and no-tool conditions are separate controls. Across five open-weight 7-9B models, tool returns strongly dominate elicited closed-book answers. The four instruction-tuned models retain a verified-correct answer against an incorrect tool in only 6.5-17.1% of eligible cases, follow a correct tool in 86.0-93.1%, and repeat the tool return in 78.4-86.0% of cases where both sources are wrong. No cross-model ordering remains stable across three instruction-wording variants with the question and episode content held fixed. We compare prompting with SFT and DPO using chain-level cross-fitting over ToolHop, so questions sharing an underlying fact never straddle training and evaluation. We apply an asymmetric success criterion: correct-answer retention must improve without a detected reduction in correct-tool following. SFT and DPO meet this criterion on the same two of four instruction-tuned backbones. Improvements rarely come cleanly: 19 of 20 tested method-model combinations reduce abstention after tool errors or on unanswerable inputs. Transfer beyond MemToC is positive but partial and depends on the model and presentation frame. Correctness-conditioned arbitration can be improved through fine-tuning, but gains must be evaluated jointly with correct tool use, abstention, and robustness to formulation.