benchmarks evals
Live-State Wrappers Cut Grounding Failures About 78%
A 200-session game-coaching benchmark separated correct answers from correct direction.
Published Updated Story ID: mp-2026-09-24-016
Summary
A 200-session game-coaching benchmark separated correct answers from correct direction.
State-grounded wrappers lifted session-level grounded accuracy to 83.5% from 20.0% for prompting and 26.5% for a tool-use baseline. The authors do not establish independent effects for each wrapper.
Why it matters
A 200-session game-coaching benchmark separated correct answers from correct direction.
Limits and context
- The authors do not establish independent effects for each wrapper.
Key claims
A 200-session game-coaching benchmark separated correct answers from correct direction.
Qualification: The authors do not establish independent effects for each wrapper.
Evidence: source-2026-09-24-018
Sources
- arXiv preprint 2609.27606arXiv · primary research
Corrections
No corrections have been recorded for this story.