We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
1 parent 4ff2186 commit 61c343bCopy full SHA for 61c343b
1 file changed
run_paper_figures.py
@@ -1253,8 +1253,8 @@ def build_dataset_stats_table(
1253
"""
1254
caption = (
1255
r"Dataset statistics per (model, dataset) combination. "
1256
- r"\#Traj.\ counts all agent runs including those that hit the step limit. "
1257
- r"\#Tokens is the total token count across all runs. "
+ r"\#Traj.\ counts all agent runs by each model on each dataset. "
+ r"\#Tokens is the sum of tokens across all linearized trajectories. "
1258
r"$\geq$15 and $\geq$50 count trajectories reaching those turn thresholds, "
1259
r"the length filters used in the lookahead experiments. "
1260
r"The $k_{\max}=15$ filter is nearly lossless (90--99\,\% of runs); "
0 commit comments