# Unable to Forget: Proactive Interference Reveals Working Memory Limits in LLMs Beyond Context Length

arXiv: https://arxiv.org/abs/2506.08184 · HTML: https://mlsanity.com/paper/2506.08184/

## Key facts

- Cited by 20 papers in the map (data through 14 Oct 2026).
- Builds on 4 of its 14 references that are in the map.
- First posted on arXiv 9 Jun 2025.

## Builds on (most cited first, 4 of 4)

- [Lost in the Middle: How Language Models Use Long Contexts](https://mlsanity.com/paper/2307.03172/) (2023)
- [Does Prompt Formatting Have Any Impact on LLM Performance?](https://mlsanity.com/paper/2411.10541/) (2024)
- [Michelangelo: Long Context Evaluations Beyond Haystacks via Latent Structure Queries](https://mlsanity.com/paper/2409.12640/) (2024)
- [Working Memory Capacity of ChatGPT: An Empirical Study](https://mlsanity.com/paper/2305.03731/) (2023)

## Built on by (most cited first, 20 of 20)

- [Large Language Model Reasoning Failures](https://mlsanity.com/paper/2602.06176/) (2026)
- [Sculptor: Empowering LLMs with Cognitive Agency via Active Context Management](https://mlsanity.com/paper/2508.04664/) (2025)
- [COMI: Coarse-to-fine Context Compression via Marginal Information Gain](https://mlsanity.com/paper/2602.01719/) (2026)
- [Learning to Forget: Sleep-Inspired Memory Consolidation for Resolving Proactive Interference in Large Language Models](https://mlsanity.com/paper/2603.14517/) (2026)
- [A Survey on AgentOps: Categorization, Challenges, and Future Directions](https://mlsanity.com/paper/2508.02121/) (2025)
- [WMF-AM: Probing LLM Working Memory via Depth-Parameterized Cumulative State Tracking](https://mlsanity.com/paper/2603.27343/) (2026)
- [PISA: A Pragmatic Psych-Inspired Unified Memory System for Enhanced AI Agency](https://mlsanity.com/paper/2510.15966/) (2025)
- [SinkTrack: Attention Sink based Context Anchoring for Large Language Models](https://mlsanity.com/paper/2604.10027/) (2026)
- [LLMs Remember First, Forget Last: Dual-Process Interference in Large Language Models](https://mlsanity.com/paper/2603.00270/) (2026)
- [How Task Structure Limits Multi-Agent Success: An Information-Theoretic Analysis](https://mlsanity.com/paper/2606.13733/) (2026)
- [Compress and Forget: bitsandbytes Quantization Amplifies Proactive Interference in LLMs](https://mlsanity.com/paper/2608.18578/) (2026)
- [A New Strategy for Artificial Intelligence: Training Foundation Models Directly on Human Brain Data](https://mlsanity.com/paper/2601.12053/) (2026)
- [Prompt Complexity Dilutes Structured Reasoning: A Follow-Up Study on the Car Wash Problem](https://mlsanity.com/paper/2603.13351/) (2026)
- [Agent System Operations: Categorization, Challenges, and Future Directions](https://mlsanity.com/paper/2606.01581/) (2026)
- [Instruction Bleed: Cross-Module Interference in Prompt-Composed Agentic Systems](https://mlsanity.com/paper/2606.26356/) (2026)
- [The Sleeping Agent: What Gist-Based Context Compression Loses and Why](https://mlsanity.com/paper/2608.11775/) (2026)
- [Self-Cleaning and Captured Anyway: One Measured Primitive for Error in a Store an Agent Writes to Itself, and What a Falling Score Actually Measures](https://mlsanity.com/paper/2609.25052/) (2026)
- [In-Context Binding Capacity in Language Models](https://mlsanity.com/paper/2609.30634/) (2026)
- [Positions Are Not Facts: The Mismatch Between KV Caches and Memory](https://mlsanity.com/paper/2609.33759/) (2026)
- [When Context Changes: Understanding Update Failures in LLMs](https://mlsanity.com/paper/2609.38866/) (2026)

## Abstract (arXiv)

Information retrieval in Large Language Models (LLMs) is increasingly recognized as intertwined with generation capabilities rather than mere lookup. While longer contexts are often assumed to improve retrieval, the effects of intra-context interference remain understudied. To address this, we adapt the proactive interference (PI) paradigm from cognitive science, where earlier information disrupts recall of newer updates. In humans, susceptibility to such interference is inversely linked to working memory capacity. We introduce PI-LLM, an evaluation that sequentially streams semantically related key-value updates and queries only the final values. Although these final values are clearly positioned just before the query, LLM retrieval accuracy declines log-linearly toward zero as interference accumulates; errors arise from retrieving previously overwritten values. Attempts to mitigate interference via prompt engineering (e.g., instructing models to ignore earlier input) yield limited success. These findings reveal a fundamental constraint on LLMs' ability to disentangle interference and flexibly manipulate information, suggesting a working memory bottleneck beyond mere context access. This calls for approaches that strengthen models' ability to suppress irrelevant content during retrieval.