Yesterday
-
Remove superseded Qwen results and medium-reasoning Sol pilot
-
Show measured efficiency values with incomplete coverage markers
-
Publish reconstruction readouts and benchmark leaderboard
-
Backfill retained Sol usage and pricing-derived inference costs
-
Publish reconstruction readouts and benchmark leaderboard
-
Add paired viewer images, confidence intervals and leaderboard plots
-
Add paired viewer images, confidence intervals and leaderboard plots
-
Publish reconstruction readouts and benchmark leaderboard
2 days ago
-
Update reconstruction readouts and score viewer
-
Retire the superseded 32-turn comparison
-
Update reconstruction readouts and score viewer
-
Clear previous benchmark results
-
Update the GPT-6.1-Sol readout and score viewer to 5 reconstructions
