E1 · Stress set: harder scenarios on the frozen code (no change to what the model sees); sol and terra, experiments/stress1. Not part of the headline.

optobench runs

48 trials · total cost $12.48 · the 8 Claude Fable and Opus E1 episodes (stress1_anthropic, $34.36) are not on this page; they are in the dataset's episodes table and the Results page · click a trial to step through it

By agent and task (trials shown by the filters)

agent task trialsharness errors submitted all passmean end score costcost / trial
solt1_connectivity12—1210 / 120.99$3.9510$0.3293
solt3_light3—33 / 30.944$1.2496$0.4165
solt5_traces9—99 / 90.998$2.3814$0.2646
solall tasks24—2422 / 240.987$7.5820$0.3159
terrat1_connectivity12—127 / 120.982$2.3970$0.1998
terrat3_light3—33 / 30.944$0.7886$0.2629
terrat5_traces9—99 / 90.997$1.7114$0.1902
terraall tasks24—2419 / 240.983$4.8970$0.2040

Each trial has one end grade, computed once after the run ends. Harness failures (the run stopped on a provider or harness error) are listed but have no score: they are left out of "all pass" and the means.

Trials

headlinesolt1_connectivityloose_wiretrial 0submitted1pass$0.25963148
headlinesolt1_connectivityloose_wiretrial 1submitted1pass$0.25042849
headlinesolt1_connectivityloose_wiretrial 2submitted1pass$0.37993860
headlinesolt1_connectivitymask_fliptrial 0submitted1pass$0.32624162
headlinesolt1_connectivitymask_fliptrial 1submitted0.957fail$0.44544872
headlinesolt1_connectivitymask_fliptrial 2submitted1pass$0.37874366
headlinesolt1_connectivityred_led_agedtrial 0submitted1pass$0.27332846
headlinesolt1_connectivityred_led_agedtrial 1submitted1pass$0.30663657
headlinesolt1_connectivityred_led_agedtrial 2submitted0.923fail$0.34233659
headlinesolt1_connectivitystale_framestrial 0submitted1pass$0.31463759
headlinesolt1_connectivitystale_framestrial 1submitted1pass$0.36763159
headlinesolt1_connectivitystale_framestrial 2submitted1pass$0.30643455
headlineterrat1_connectivityloose_wiretrial 0submitted1pass$0.12462642
headlineterrat1_connectivityloose_wiretrial 1submitted1pass$0.14882643
headlineterrat1_connectivityloose_wiretrial 2submitted1pass$0.16422742
headlineterrat1_connectivitymask_fliptrial 0submitted0.888fail$0.22373351
headlineterrat1_connectivitymask_fliptrial 1submitted1pass$0.18554162
headlineterrat1_connectivitymask_fliptrial 2submitted0.969fail$0.22223960
headlineterrat1_connectivityred_led_agedtrial 0submitted0.927fail$0.26453661
headlineterrat1_connectivityred_led_agedtrial 1submitted1pass$0.14962238
headlineterrat1_connectivityred_led_agedtrial 2submitted1pass$0.24173151
headlineterrat1_connectivitystale_framestrial 0submitted1pass$0.17502646
headlineterrat1_connectivitystale_framestrial 1submitted0.998fail$0.22933056
headlineterrat1_connectivitystale_framestrial 2submitted0.998fail$0.26792962
headlinesolt3_lightconverging_led_lowtrial 0submitted0.973pass$0.38725682
headlinesolt3_lightconverging_led_lowtrial 1submitted0.929pass$0.38154376
headlinesolt3_lightconverging_led_lowtrial 2submitted0.929pass$0.48085893
headlineterrat3_lightconverging_led_lowtrial 0submitted0.973pass$0.28126199
headlineterrat3_lightconverging_led_lowtrial 1submitted0.929pass$0.23873165
headlineterrat3_lightconverging_led_lowtrial 2submitted0.929pass$0.26865893
headlinesolt5_tracesbleached_and_droppedtrial 0submitted0.995pass$0.2658124
headlinesolt5_tracesbleached_and_droppedtrial 1submitted0.999pass$0.2089120
headlinesolt5_tracesbleached_and_droppedtrial 2submitted0.999pass$0.2789124
headlinesolt5_traceseverythingtrial 0submitted0.998pass$0.2036118
headlinesolt5_traceseverythingtrial 1submitted0.997pass$0.2064120
headlinesolt5_traceseverythingtrial 2submitted0.993pass$0.5219133
headlinesolt5_traceslight_and_frozentrial 0submitted1pass$0.2439123
headlinesolt5_traceslight_and_frozentrial 1submitted1pass$0.2535118
headlinesolt5_traceslight_and_frozentrial 2submitted0.999pass$0.1985117
headlineterrat5_tracesbleached_and_droppedtrial 0submitted0.998pass$0.2098131
headlineterrat5_tracesbleached_and_droppedtrial 1submitted1pass$0.1944122
headlineterrat5_tracesbleached_and_droppedtrial 2submitted0.999pass$0.2622131
headlineterrat5_traceseverythingtrial 0submitted0.992pass$0.1853125
headlineterrat5_traceseverythingtrial 1submitted0.994pass$0.1980127
headlineterrat5_traceseverythingtrial 2submitted0.996pass$0.1805125
headlineterrat5_traceslight_and_frozentrial 0submitted0.999pass$0.1485122
headlineterrat5_traceslight_and_frozentrial 1submitted0.997pass$0.1715125
headlineterrat5_traceslight_and_frozentrial 2submitted0.998pass$0.1610125