Skip to content

Commit 81ee924

Browse files
committed
fix(evals): use result filenames in cross verification
1 parent 3d15cec commit 81ee924

2 files changed

Lines changed: 3 additions & 3 deletions

File tree

scripts/cross-verify-parallel.sh

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -25,7 +25,7 @@ run_one() {
2525
local dir="$1"
2626
local approach="$2"
2727
local label="cross-${approach}"
28-
local out_file="$dir/scores/mmrubric_${label}.json"
28+
local out_file="$dir/scores/result_${label}.json"
2929
local task
3030
task=$(basename "$dir")
3131
if [[ -f "$out_file" ]]; then

scripts/cross-verify.sh

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -3,7 +3,7 @@
33
# Lets us isolate verifier disagreement from agent variance.
44
#
55
# Inputs: every trajectory dir matched by TRAJECTORY_GLOB.
6-
# Outputs: scores/mmrubric_cross-{outcome-only,a,b}.json next to each trajectory.
6+
# Outputs: scores/result_cross-{outcome-only,a,b}.json next to each trajectory.
77

88
set -e
99
cd "$(dirname "$0")/.."
@@ -27,7 +27,7 @@ for d in "${DIRS[@]}"; do
2727
echo "=== $(basename "$(dirname "$d")")/$task ==="
2828
for approach in outcome-only b a; do
2929
label="cross-${approach}"
30-
out_file="$d/scores/mmrubric_${label}.json"
30+
out_file="$d/scores/result_${label}.json"
3131
if [[ -f "$out_file" ]]; then
3232
echo " [$approach] already exists, skipping"
3333
continue

0 commit comments

Comments
 (0)