harness/round2_windows.csv

Sixty-second windows: every audio and text model vs the enumerator, with baseline and human floor

39 rows · raw file: harness/round2_windows.csv

armmodelsubtaskncount_rMAEbiaswithin5word_precisionword_recallbaseline_rbaseline_MAEhuman_floor_agreecost_per_window
audiogoogle/gemini-3.1-pro-previewidwrd_urd330.991.70.1970.750.790.966.650.80.0566
audioopenai/gpt-audioidwrd_urd120.933.31.3920.320.440.986.850.80.0294
audiogoogle/gemini-3.8-flashidwrd_urd1490.933.80.2860.680.590.869.050.80.0098
audiogoogle/gemini-3.5-flash-liteidwrd_urd350.768.0-0.9690.260.520.98.650.80.0032
audioopenai/gpt-audio-miniidwrd_urd350.611.9-2.1570.190.640.98.650.80.003
audioxiaomi/mimo-v2.5idwrd_urd340.611.97.3620.250.310.98.650.80.0011
audiogoogle/gemini-2.5-flashidwrd_urd350.5612.810.9570.30.170.98.650.80.0043
audionvidia/nemotron-3-nano-omni-30b-a3b-reasoning:freeidwrd_urd150.5215.315.3470.00.00.948.550.80.0
audiomistralai/voxtral-small-24b-2507idwrd_urd1022.822.8400.00.00.949.550.80.0065
textanthropic/claude-sonnet-5idwrd_urd270.934.2-3.7890.460.830.946.150.80.0379
textgoogle/gemini-3-flash-previewidwrd_urd350.96.7-6.1630.310.860.98.650.80.0025
textopenai/gpt-5-miniidwrd_urd340.818.9-8.4650.230.850.928.050.80.0077
textdeepseek/deepseek-v3.2idwrd_urd350.8611.5-11.1400.20.920.98.650.80.0008
audioopenai/gpt-audioorf_eng110.944.2-4.0730.180.330.7113.861.50.0293
audiogoogle/gemini-3.5-flash-liteorf_eng250.955.1-2.4720.250.380.8611.361.50.0033
audiogoogle/gemini-3.8-flashorf_eng1040.895.8-0.2740.310.380.811.261.50.0092
audiogoogle/gemini-3.1-pro-previeworf_eng240.946.1-1.5710.190.310.8810.461.50.0573
audiogoogle/gemini-2.5-flashorf_eng250.867.75.2680.220.130.8611.361.50.0046
audioxiaomi/mimo-v2.5orf_eng230.829.14.2520.290.280.8511.761.50.0011
audionvidia/nemotron-3-nano-omni-30b-a3b-reasoning:freeorf_eng1221.521.3420.00.00.8810.761.50.0
audiomistralai/voxtral-small-24b-2507orf_eng1022.222.0300.00.00.8912.861.50.0065
audioopenai/gpt-audio-miniorf_eng24-0.0531.7-17.0290.120.770.8511.661.50.0033
textanthropic/claude-sonnet-5orf_eng200.869.2-8.4550.20.420.889.261.50.033
textgoogle/gemini-3-flash-previeworf_eng250.8611.3-10.8440.180.490.8611.361.50.0027
textdeepseek/deepseek-v3.2orf_eng230.8411.6-10.7430.190.480.8511.761.50.0016
textopenai/gpt-5-miniorf_eng240.4720.7-19.0420.160.770.8511.361.50.0082
audiogoogle/gemini-3.1-pro-previeworf_urd280.972.6-2.0890.590.810.9315.554.30.0551
audiogoogle/gemini-3.8-flashorf_urd1290.95.5-1.2740.480.520.8715.754.30.0097
audiogoogle/gemini-3.5-flash-liteorf_urd340.847.0-0.1620.390.680.9115.454.30.0037
audioopenai/gpt-audioorf_urd150.6413.1-9.2270.130.810.920.054.30.0302
audiogoogle/gemini-2.5-flashorf_urd340.4913.610.1620.570.250.9115.454.30.0047
audioxiaomi/mimo-v2.5orf_urd330.5113.87.7580.180.150.9115.654.30.0014
audionvidia/nemotron-3-nano-omni-30b-a3b-reasoning:freeorf_urd1517.817.8530.00.00.9417.154.30.0
audiomistralai/voxtral-small-24b-2507orf_urd1120.020.0450.00.00.9117.854.30.0065
audioopenai/gpt-audio-miniorf_urd340.1525.9-11.1290.10.720.9115.454.30.0036
textanthropic/claude-sonnet-5orf_urd240.958.5-8.1460.190.780.9712.654.30.0418
textopenai/gpt-5-miniorf_urd310.8512.8-12.4260.180.890.9414.154.30.0081
textgoogle/gemini-3-flash-previeworf_urd340.8714.9-14.8290.160.910.9115.454.30.0029
textdeepseek/deepseek-v3.2orf_urd340.821.5-21.5240.130.970.9115.454.30.0007