===== MB round-1 task=helpsteer2 anchor=/NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_r2/train/ks weights={"helpfulness":0.2500,"correctness":0.2500,"coherence":0.2500,"conciseness":0.2500} =====
manifest 2500
allresp built
scores combined {'pol_42': 2500, 'pol_43': 2500, 'base_44': 2500, 'base_45': 2500}
{"surplus(2P-1)": {"helpfulness": 0.008, "correctness": 0.014, "coherence": 0.009, "conciseness": -0.022}, "P(pi>mu)": {"helpfulness": 0.504, "correctness": 0.507, "coherence": 0.504, "conciseness": 0.489}, "w_nbs": {"helpfulness": 0.0, "correctness": 0.0, "coherence": 0.0, "conciseness": 0.999}, "w_ks": {"helpfulness": 0.247, "correctness": 0.225, "coherence": 0.244, "conciseness": 0.284}, "worst": "conciseness", "n": 5000}
RMPOOL_helpsteer2_DONE
[build_bon] 2500 pairs (train 2250/test 250) weights {'helpfulness': 0.25, 'correctness': 0.25, 'coherence': 0.25, 'conciseness': 0.25}
[prep helpsteer2 EXP=exp/rmpool_mbhtconc431] objs: helpfulness correctness coherence conciseness parent: /NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_r2/train/ks

Map:   0%|          | 0/2250 [00:00<?, ? examples/s]
Map:  89%|████████▉ | 2000/2250 [00:00<00:00, 17972.96 examples/s]
Map: 100%|██████████| 2250/2250 [00:00<00:00, 17348.88 examples/s]

Map:   0%|          | 0/250 [00:00<?, ? examples/s]
Map: 100%|██████████| 250/250 [00:00<00:00, 13858.14 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/2250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 214720.58 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 213033.79 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 51664.17 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 50886.93 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/2250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 296813.46 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 293828.51 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 60377.50 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 59355.60 examples/s]
[prep] pre_pref + pre_htobj built
dpo htmnpo_coherence htmnpo_conciseness htmnpo_correctness htmnpo_helpfulness inpo ipo ks maxmin nbs simpo unif 
TRAIN_DONE aiprlab-polymer
===== MB round-2 task=helpsteer2 anchor=/NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtconc431/train/htmnpo_conciseness weights={"helpfulness":0.2500,"correctness":0.2500,"coherence":0.2500,"conciseness":0.2500} =====
manifest 2500
allresp built
scores combined {'pol_42': 2500, 'pol_43': 2500, 'base_44': 2500, 'base_45': 2500}
{"surplus(2P-1)": {"helpfulness": 0.003, "correctness": 0.002, "coherence": 0.002, "conciseness": 0.001}, "P(pi>mu)": {"helpfulness": 0.501, "correctness": 0.501, "coherence": 0.501, "conciseness": 0.5}, "w_nbs": {"helpfulness": 0.167, "correctness": 0.181, "coherence": 0.217, "conciseness": 0.435}, "w_ks": {"helpfulness": 0.247, "correctness": 0.248, "coherence": 0.25, "conciseness": 0.254}, "worst": "conciseness", "n": 5000}
RMPOOL_helpsteer2_DONE
[build_bon] 2500 pairs (train 2250/test 250) weights {'helpfulness': 0.25, 'correctness': 0.25, 'coherence': 0.25, 'conciseness': 0.25}
[prep helpsteer2 EXP=exp/rmpool_mbhtconc432] objs: helpfulness correctness coherence conciseness parent: /NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtconc431/train/htmnpo_conciseness

Map:   0%|          | 0/2250 [00:00<?, ? examples/s]
Map:  89%|████████▉ | 2000/2250 [00:00<00:00, 17077.47 examples/s]
Map: 100%|██████████| 2250/2250 [00:00<00:00, 16611.19 examples/s]

Map:   0%|          | 0/250 [00:00<?, ? examples/s]
Map: 100%|██████████| 250/250 [00:00<00:00, 9811.15 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/2250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 186229.58 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 183935.60 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 50746.55 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 49901.30 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/2250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 324725.90 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 321047.25 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 59520.69 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 58511.02 examples/s]
[prep] pre_pref + pre_htobj built
dpo htmnpo_coherence htmnpo_conciseness htmnpo_correctness htmnpo_helpfulness inpo ipo ks maxmin nbs simpo unif 
TRAIN_DONE aiprlab-polymer
===== MB round-3 task=helpsteer2 anchor=/NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtconc432/train/htmnpo_conciseness weights={"helpfulness":0.2500,"correctness":0.2500,"coherence":0.2500,"conciseness":0.2500} =====
manifest 2500
allresp built
scores combined {'pol_42': 2500, 'pol_43': 2500, 'base_44': 2500, 'base_45': 2500}
{"surplus(2P-1)": {"helpfulness": -0.008, "correctness": -0.01, "coherence": -0.021, "conciseness": -0.011}, "P(pi>mu)": {"helpfulness": 0.496, "correctness": 0.495, "coherence": 0.489, "conciseness": 0.494}, "w_nbs": {"helpfulness": 0.25, "correctness": 0.25, "coherence": 0.25, "conciseness": 0.25}, "w_ks": {"helpfulness": 0.25, "correctness": 0.25, "coherence": 0.25, "conciseness": 0.25}, "worst": "coherence", "n": 5000}
RMPOOL_helpsteer2_DONE
[build_bon] 2500 pairs (train 2250/test 250) weights {'helpfulness': 0.25, 'correctness': 0.25, 'coherence': 0.25, 'conciseness': 0.25}
[prep helpsteer2 EXP=exp/rmpool_mbhtconc433] objs: helpfulness correctness coherence conciseness parent: /NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtconc432/train/htmnpo_conciseness

Map:   0%|          | 0/2250 [00:00<?, ? examples/s]
Map:  89%|████████▉ | 2000/2250 [00:00<00:00, 18139.37 examples/s]
Map: 100%|██████████| 2250/2250 [00:00<00:00, 12977.98 examples/s]

Map:   0%|          | 0/250 [00:00<?, ? examples/s]
Map: 100%|██████████| 250/250 [00:00<00:00, 13473.34 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/2250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 226289.66 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 224438.36 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 50060.92 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 49333.15 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/2250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 275200.75 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 272703.69 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 61518.10 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 60440.14 examples/s]
[prep] pre_pref + pre_htobj built
dpo htmnpo_coherence htmnpo_conciseness htmnpo_correctness htmnpo_helpfulness inpo ipo ks maxmin nbs simpo unif 
TRAIN_DONE aiprlab-polymer
===== MB round-4 task=helpsteer2 anchor=/NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtconc433/train/htmnpo_conciseness weights={"helpfulness":0.2500,"correctness":0.2500,"coherence":0.2500,"conciseness":0.2500} =====
manifest 2500
allresp built
scores combined {'pol_42': 2500, 'pol_43': 2500, 'base_44': 2500, 'base_45': 2500}
{"surplus(2P-1)": {"helpfulness": -0.0, "correctness": 0.004, "coherence": -0.002, "conciseness": -0.0}, "P(pi>mu)": {"helpfulness": 0.5, "correctness": 0.502, "coherence": 0.499, "conciseness": 0.5}, "w_nbs": {"helpfulness": 0.333, "correctness": 0.0, "coherence": 0.333, "conciseness": 0.333}, "w_ks": {"helpfulness": 0.254, "correctness": 0.238, "coherence": 0.254, "conciseness": 0.254}, "worst": "coherence", "n": 5000}
RMPOOL_helpsteer2_DONE
[build_bon] 2500 pairs (train 2250/test 250) weights {'helpfulness': 0.25, 'correctness': 0.25, 'coherence': 0.25, 'conciseness': 0.25}
[prep helpsteer2 EXP=exp/rmpool_mbhtconc434] objs: helpfulness correctness coherence conciseness parent: /NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtconc433/train/htmnpo_conciseness

Map:   0%|          | 0/2250 [00:00<?, ? examples/s]
Map:  89%|████████▉ | 2000/2250 [00:00<00:00, 17707.31 examples/s]
Map: 100%|██████████| 2250/2250 [00:00<00:00, 16846.85 examples/s]

Map:   0%|          | 0/250 [00:00<?, ? examples/s]
Map: 100%|██████████| 250/250 [00:00<00:00, 12408.89 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/2250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 226740.92 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 223810.27 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 50216.75 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 49296.04 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/2250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 242041.14 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 239237.05 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 56781.07 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 55680.54 examples/s]
[prep] pre_pref + pre_htobj built
dpo htmnpo_coherence htmnpo_conciseness htmnpo_correctness htmnpo_helpfulness inpo ipo ks maxmin nbs simpo unif 
TRAIN_DONE aiprlab-polymer
===== MB round-5 task=helpsteer2 anchor=/NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtconc434/train/htmnpo_conciseness weights={"helpfulness":0.2500,"correctness":0.2500,"coherence":0.2500,"conciseness":0.2500} =====
manifest 2500
allresp built
scores combined {'pol_42': 2500, 'pol_43': 2500, 'base_44': 2500, 'base_45': 2500}
{"surplus(2P-1)": {"helpfulness": 0.007, "correctness": 0.013, "coherence": 0.007, "conciseness": -0.013}, "P(pi>mu)": {"helpfulness": 0.504, "correctness": 0.506, "coherence": 0.503, "conciseness": 0.494}, "w_nbs": {"helpfulness": 0.0, "correctness": 0.0, "coherence": 0.0, "conciseness": 0.999}, "w_ks": {"helpfulness": 0.247, "correctness": 0.226, "coherence": 0.248, "conciseness": 0.279}, "worst": "conciseness", "n": 5000}
RMPOOL_helpsteer2_DONE
[build_bon] 2500 pairs (train 2250/test 250) weights {'helpfulness': 0.25, 'correctness': 0.25, 'coherence': 0.25, 'conciseness': 0.25}
[prep helpsteer2 EXP=exp/rmpool_mbhtconc435] objs: helpfulness correctness coherence conciseness parent: /NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtconc434/train/htmnpo_conciseness

Map:   0%|          | 0/2250 [00:00<?, ? examples/s]
Map:  89%|████████▉ | 2000/2250 [00:00<00:00, 18387.60 examples/s]
Map: 100%|██████████| 2250/2250 [00:00<00:00, 17816.50 examples/s]

Map:   0%|          | 0/250 [00:00<?, ? examples/s]
Map: 100%|██████████| 250/250 [00:00<00:00, 13565.55 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/2250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 240376.57 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 238108.29 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 54175.98 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 53324.65 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/2250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 339040.20 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 335210.60 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 60829.33 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 59788.80 examples/s]
[prep] pre_pref + pre_htobj built
dpo htmnpo_coherence htmnpo_conciseness htmnpo_correctness htmnpo_helpfulness inpo ipo ks maxmin nbs simpo unif 
TRAIN_DONE aiprlab-polymer
===== MB round-6 task=helpsteer2 anchor=/NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtconc435/train/htmnpo_conciseness weights={"helpfulness":0.2500,"correctness":0.2500,"coherence":0.2500,"conciseness":0.2500} =====
manifest 2500
allresp built
scores combined {'pol_42': 2500, 'pol_43': 2500, 'base_44': 2500, 'base_45': 2500}
{"surplus(2P-1)": {"helpfulness": 0.007, "correctness": -0.004, "coherence": -0.001, "conciseness": -0.018}, "P(pi>mu)": {"helpfulness": 0.503, "correctness": 0.498, "coherence": 0.499, "conciseness": 0.491}, "w_nbs": {"helpfulness": 0.0, "correctness": 0.333, "coherence": 0.333, "conciseness": 0.333}, "w_ks": {"helpfulness": 0.23, "correctness": 0.257, "coherence": 0.257, "conciseness": 0.257}, "worst": "conciseness", "n": 5000}
RMPOOL_helpsteer2_DONE
[build_bon] 2500 pairs (train 2250/test 250) weights {'helpfulness': 0.25, 'correctness': 0.25, 'coherence': 0.25, 'conciseness': 0.25}
[prep helpsteer2 EXP=exp/rmpool_mbhtconc436] objs: helpfulness correctness coherence conciseness parent: /NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtconc435/train/htmnpo_conciseness

Map:   0%|          | 0/2250 [00:00<?, ? examples/s]
Map:  89%|████████▉ | 2008/2250 [00:00<00:00, 18907.22 examples/s]
Map: 100%|██████████| 2250/2250 [00:00<00:00, 18041.78 examples/s]

Map:   0%|          | 0/250 [00:00<?, ? examples/s]
Map: 100%|██████████| 250/250 [00:00<00:00, 14029.84 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/2250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 213158.90 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 210140.15 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 51623.47 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 50487.55 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/2250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 342696.78 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 338541.54 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 60741.24 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 59679.91 examples/s]
[prep] pre_pref + pre_htobj built
dpo htmnpo_coherence htmnpo_conciseness htmnpo_correctness htmnpo_helpfulness inpo ipo ks maxmin nbs simpo unif 
TRAIN_DONE aiprlab-polymer
===== MB round-7 task=helpsteer2 anchor=/NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtconc436/train/htmnpo_conciseness weights={"helpfulness":0.2500,"correctness":0.2500,"coherence":0.2500,"conciseness":0.2500} =====
manifest 2500
allresp built
scores combined {'pol_42': 2500, 'pol_43': 2500, 'base_44': 2500, 'base_45': 2500}
{"surplus(2P-1)": {"helpfulness": 0.006, "correctness": -0.004, "coherence": 0.01, "conciseness": -0.006}, "P(pi>mu)": {"helpfulness": 0.503, "correctness": 0.498, "coherence": 0.505, "conciseness": 0.497}, "w_nbs": {"helpfulness": 0.0, "correctness": 0.5, "coherence": 0.0, "conciseness": 0.5}, "w_ks": {"helpfulness": 0.24, "correctness": 0.267, "coherence": 0.226, "conciseness": 0.267}, "worst": "conciseness", "n": 5000}
RMPOOL_helpsteer2_DONE
[build_bon] 2500 pairs (train 2250/test 250) weights {'helpfulness': 0.25, 'correctness': 0.25, 'coherence': 0.25, 'conciseness': 0.25}
[prep helpsteer2 EXP=exp/rmpool_mbhtconc437] objs: helpfulness correctness coherence conciseness parent: /NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtconc436/train/htmnpo_conciseness

Map:   0%|          | 0/2250 [00:00<?, ? examples/s]
Map:  89%|████████▉ | 2005/2250 [00:00<00:00, 18787.91 examples/s]
Map: 100%|██████████| 2250/2250 [00:00<00:00, 17844.59 examples/s]

Map:   0%|          | 0/250 [00:00<?, ? examples/s]
Map: 100%|██████████| 250/250 [00:00<00:00, 13440.70 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/2250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 252472.89 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 250117.52 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 54451.68 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 53547.95 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/2250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 347799.22 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 343682.73 examples/s]

Saving the dataset (0/1 shards):   0%|          | 0/250 [00:00<?, ? examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 61013.38 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 59949.46 examples/s]
[prep] pre_pref + pre_htobj built
dpo htmnpo_coherence htmnpo_conciseness htmnpo_correctness htmnpo_helpfulness inpo ipo ks maxmin nbs simpo unif 
TRAIN_DONE aiprlab-polymer
{
  "label": "qwen32_helpsteer2_htmnpo_conciseness",
  "n_objectives": 4,
  "surplus": {
    "helpfulness": -0.2685,
    "correctness": -0.1832,
    "coherence": -0.2049,
    "conciseness": 0.3234
  },
  "hoeffding_margin": {
    "helpfulness": 0.0356,
    "correctness": 0.0358,
    "coherence": 0.0356,
    "conciseness": 0.0356
  },
  "worst_surplus": -0.2685,
  "individual_rationality": false,
  "nash_welfare_logsum": -42.5755
}
EVAL_helpsteer2_DONE aiprlab-polymer
MB_STAGES_DONE