cs.CLNov 15, 2026

The Silent Cost of Conversation History in LLM Reasoning

Authors: Ali Şenol

Organizations: Department of Computer Engineering, Tarsus University, Tarsus, Mersin, 33400, Türkiye

Abstract

Large language models (LLMs) in multi-turn information systems accumulate their own earlier replies and a growing task state. We asked whether this conversation history lowers accuracy, and whether process-level quality metrics warn before it does. Each analysis was fixed before its outcomes were computed. In Study 1, on 48 sharded math tasks, retaining its replies lowered final success by 0.21 for GPT-4o-mini and 0.29 for Mistral-7B (Holm-adjusted p = 0.026 and 0.004); Llama-4 Scout showed the same direction, not significant. Study 2 introduces CDR-Stream, 100-turn state-tracking dialogues with an exact answer per turn and per-model difficulty calibration. With the full history, accuracy fell from about 0.6 to 0.1 within twenty turns at matched difficulty, whereas a solver-verified state summary kept it 0.52-0.59 higher (Holm-adjusted p less than 0.001). Consistency and logical coherence fell far less than accuracy, and stability rose. None of these metrics, a paired readiness measure or calibrated single-stream drift detectors gave robust, model-general early warning, although in Study 2 the repeat-based metrics were complete only for GPT-4o-mini. In the monitors we tested, a single 100-turn stream had power only for very large drops. Systems built on LLMs should keep task state outside the transcript and monitor across sessions.

Figures & tables

Explore similar work

CardsList
  1. What if LLMs Ate Their Words: Causal History Effects in Multi-Turn Interaction

    Sep 5, 2026Jinnan Li, Zheren Fu, Yue Wang +3LLM Context ManagementCausal Interventions in Language Models

  2. Do LLMs Benefit From Their Own Words?

    Feb 27, 2026Jenny Y. Huang, Leshem Choshen, Wei Sun +5LLM Context ManagementContext Compression