LLMs struggle to detect context switches in multi-turn dialogues, with only some reasoning models succeeding while open-weight models carry stale context and all exhibit position bias.
Similarly, including challenging examples of late-pivot detections can help reduce the degradation in performance, as context grows
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Beyond Continuity: Challenges of Context Switching in Multi-Turn Dialogue with LLMs
LLMs struggle to detect context switches in multi-turn dialogues, with only some reasoning models succeeding while open-weight models carry stale context and all exhibit position bias.