Story

arxiv_cs_ai ยท Nov 12, 2026 ยท paper

Source brief

When Both Sides Adapt: The Challenges of Evaluating Interactive LLM Systems

arxiv.orgNov 12, 2026
original source linked

In brief

In interactive systems where users engage with LLMs, users are adaptive and, after observing LLM responses, can change their prompting strategy, as well as the type and difficulty of subsequent tasks. Thus model respo...

Feed lens
evaluation

Continue reading

Read the original at arxiv.org โ†’Open in live feedRead that dayโ€™s brief

Earlier in this thread 4 items