Glossary · term93 of 202
HumanEval
A code-generation benchmark where prompts include a function signature, which lets it survive prompt compression better than benchmarks without that structural anchor.
Related terms
MBPPA benchmark of short Python programming problems, used to measure how much a compressed prompt inflates a model's output length compared to an uncompressed baseline.view termoutput expansionThe tendency of a model to generate a longer answer once a compressed prompt strips the structural cues that normally signal how terse to be.view term
At a glance
- cited by
- 2 posts
- categories
- 2
- first used
- apr 2026
Token-Compression Skills Do Not Survive MeasurementToken compression skills report savings the invoice never shows. Cache reads carry 87 percent of an agent bill, and cutting tokens can push the cost up.
Self-Hosting AI for Code: The $0/Month Developer StackBuild a complete self-hosted AI coding stack with Ollama, Continue.dev, and Open WebUI for $0/month. Includes hardware requirements, cost math, and the setup commands.