Context-compensation scaffold for LLM evaluation prompts. A short language prefix you prepend so the model discloses prior exposure, scores on quoted evidence only, and hedges on thin evidence — for scorers that can see your CLAUDE.md, memory, or session context. Backend-agnostic. Experimental: variance-reduction effect not yet measured.
evaluation scaffold ai-safety ai-agents rubric multi-turn debiasing llm prompt-engineering evals llm-evaluation ai-reliability lpci hermes-labs context-compensation language-as-state agent-scaffold drift-recovery recovery-scaffold
-
Updated
May 27, 2026 - Python