Glossary · term78 of 202
F1 score
A single accuracy metric combining precision and recall, used in some of the studies cited to show how model performance falls as context length increases.
also written as F1
Related terms
NoLiMaA long-context benchmark that removes lexical overlap between questions and answers, measuring how far model accuracy falls as input length grows past a short baseline.view termcontext rotMeasured accuracy decline as a model's input grows longer, even when the extra tokens are irrelevant, showing up well before the advertised context limit is reached.view term
At a glance
- cited by
- 2 posts
- categories
- 2
- first used
- jul 2026
Claude Is a Math Problem (And the Context Window Is the Equation)The Claude context window is an equation, not a chat. Why failed attempts anchor the output, where measured degradation starts, and when to clear and re-derive.
Why expensive models make terrible execution subagentsSubagent model selection has one costly default. Reasoning models obey their own traces 15.6% of the time, so here is where to pin model and effort.