A Conditional No-Hair Theorem for Transformers: How Language Models Forget Their Prompt
arXiv:2610.24133v1 Announce Type: new Abstract: We ask how much an autoregressive transformer retains of its prompt after generating $t$ tokens. Consider two different prompts followed by the same continuation text, so that any difference in their hidden states comes from the prompt alone, and we f…
Read the full story at arXiv cs.AI ↗
Timeline · 1 report
- 2026-10-20 04:00 · arXiv cs.AI
A Conditional No-Hair Theorem for Transformers: How Language Models Forget Their Prompt