Why do AI assistants get worse at longer conversations?

Post angle for Medium/LinkedIn

Your AI assistant is getting dumber the longer you talk to it — and it's because we trained it to be too helpful.

That's the counterintuitive finding from two converging research papers. When LLMs receive fully-specified instructions in a single message, they perform at ~90% accuracy. But spread those same instructions across a natural conversation — revealing details gradually, the way humans actually communicate — and performance drops to ~65%. A 25-point gap. And it appears even in two-turn conversations.

What goes wrong:

LLMs make premature assumptions when information is incomplete, propose solutions too early, and then lock in to those initial guesses. When the user provides more details that contradict the early assumptions, the models can't course-correct — they get lost and don't recover.

Why it happens:

This isn't a model limitation. The Intent Mismatch paper argues it's a rational strategy induced by RLHF training. Models are trained to be helpful. Under uncertainty, being helpful means guessing rather than asking. The training literally rewards premature commitment.

The real bottleneck is pragmatic mismatch: users exhibit individual variation in how they express intent. The same fragmentary utterance might be a confirmation, a correction, or a refinement — but models aligned to the "average" user default to interpreting it as confirmation of their own assumptions.

What fixes it:

• Mediator-Assistant architecture: decouple intent understanding from task execution; a Mediator explicates latent user intent before passing to the execution Assistant
• Multi-turn-aware rewards: train for long-term interaction quality, not single-turn helpfulness
• Recapitulation: restating all revealed information periodically recovers 15-20% of lost performance — partial but insufficient
• Selective history retrieval: since Does including all conversation history actually help retrieval?, not all conversation history is equal — topic switches within sessions inject irrelevant context. Selectively retrieving relevant prior turns rather than dumping the full history addresses one mechanism of the wrong-turn cascade

The deeper point:

We built AI that's spectacular at answering questions and terrible at having conversations. The multi-turn case is the real-world case — and the training signals that made models impressive in benchmarks are the same signals that make them fragile in dialogue.

Key sources:

• Why do language models fail in gradually revealed conversations?
• Why do language models lose performance in longer conversations?
• Why do language models respond passively instead of asking clarifying questions?
• Does preference optimization harm conversational understanding?
• Why can't advanced AI models take initiative in conversation?
• Does including all conversation history actually help retrieval? — selective context manages the irrelevant-history mechanism of wrong turns