Files
trolley-bench/results/meta/checks/strands-decider-2b-reading-probe.txt
zachlandesandGitHub 462aaf404e feat: add v2 benchmark and registered reproduction sample (#1)
* Kept the original raw answers while making later runs reproducible.
* Registered a variance-sized sample before the independent rerun.
* Removed article drafts and private paths from the current tree.
2026-10-14 15:36:06 -07:00

5 lines
478 B
Plaintext

Factual questions about three v1 switch prompts (wording c1, five first), one at a time, after the grids
switch|c1|o1|0 {'main_track_empty': 0.847, 'more_than_three_on_main': 0.245, 'pulling_saves_someone': 0.322, 'pull': 0.355}
switch|c1|o1|5 {'main_track_empty': 0.256, 'more_than_three_on_main': 0.813, 'pulling_saves_someone': 0.441, 'pull': 0.481}
switch|c1|o1|100 {'main_track_empty': 0.257, 'more_than_three_on_main': 0.777, 'pulling_saves_someone': 0.447, 'pull': 0.52}