[Fix] Validate per-request benchmark errors correctly

experiments/pro6000/kimi3_pro6000_sglang_sm120_flashinfer_mxfp4/run_pr_representative_point.sh

@@ -307,13 +307,17 @@ run_bench() {
     --warmup-requests "${WARMUP_REQUESTS}" --output-file "$output" \
     --output-details --disable-tqdm \
     >"${RESULT_ROOT}/bench/${stem}.log" 2>&1
-  python3 - "$output" "${NUM_PROMPTS}" <<'PY'
+  python3 - "$output" "${NUM_PROMPTS}" "${OUTPUT_LEN}" <<'PY'
 import json, sys
-path, expected = sys.argv[1], int(sys.argv[2])
+path, expected, expected_output_len = sys.argv[1], int(sys.argv[2]), int(sys.argv[3])
 rows = [json.loads(x) for x in open(path, encoding="utf-8") if x.strip()]
 assert len(rows) == 1, (path, len(rows))
 assert rows[0].get("completed") == expected, rows[0].get("completed")
-assert not rows[0].get("errors"), rows[0].get("errors")
+errors = rows[0].get("errors") or []
+assert all(not error for error in errors), errors
+output_lens = rows[0].get("output_lens") or []
+assert len(output_lens) == expected, (len(output_lens), expected)
+assert all(length == expected_output_len for length in output_lens), output_lens
 PY
 }