"Context Poisoning": Correcting LLM Mistakes in Long Chats Can Backfire

ClickOk5811 · reddit · 2026-08-08

Correcting an LLM's mistake in a long conversation might actually reinforce the error. The back-and-forth tokens surrounding the refuted claim can inadvertently increase its weight, making the model treat it as an established fact rather than a one-off error.

This implies that repeatedly trying to explain things in the same session could be structurally worse than starting fresh with just the correct information.

Original post →

More from Models

Models channel →