===== 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_mbhtcohe431] 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, 17775.64 examples/s]
Map: 100%|██████████| 2250/2250 [00:00<00:00, 17234.38 examples/s]

Map:   0%|          | 0/250 [00:00<?, ? examples/s]
Map: 100%|██████████| 250/250 [00:00<00:00, 13765.72 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, 202653.84 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 201189.25 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, 52557.57 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 51753.42 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, 292362.96 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 289537.46 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, 60762.36 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 59727.50 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_mbhtcohe431/train/htmnpo_coherence 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": -0.031, "correctness": -0.013, "coherence": -0.027, "conciseness": 0.014}, "P(pi>mu)": {"helpfulness": 0.484, "correctness": 0.494, "coherence": 0.487, "conciseness": 0.507}, "w_nbs": {"helpfulness": 0.333, "correctness": 0.333, "coherence": 0.333, "conciseness": 0.0}, "w_ks": {"helpfulness": 0.264, "correctness": 0.264, "coherence": 0.264, "conciseness": 0.209}, "worst": "helpfulness", "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_mbhtcohe432] objs: helpfulness correctness coherence conciseness parent: /NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtcohe431/train/htmnpo_coherence

Map:   0%|          | 0/2250 [00:00<?, ? examples/s]
Map:  89%|████████▉ | 2000/2250 [00:00<00:00, 16600.32 examples/s]
Map: 100%|██████████| 2250/2250 [00:00<00:00, 15998.05 examples/s]

Map:   0%|          | 0/250 [00:00<?, ? examples/s]
Map: 100%|██████████| 250/250 [00:00<00:00, 12986.10 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, 203628.96 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 201942.65 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, 39551.00 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 38594.57 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, 228000.87 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 225209.62 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, 55736.78 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 54457.34 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_mbhtcohe432/train/htmnpo_coherence 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": -0.001, "correctness": 0.007, "coherence": -0.03, "conciseness": -0.008}, "P(pi>mu)": {"helpfulness": 0.5, "correctness": 0.504, "coherence": 0.485, "conciseness": 0.496}, "w_nbs": {"helpfulness": 0.333, "correctness": 0.0, "coherence": 0.333, "conciseness": 0.333}, "w_ks": {"helpfulness": 0.257, "correctness": 0.228, "coherence": 0.257, "conciseness": 0.257}, "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_mbhtcohe433] objs: helpfulness correctness coherence conciseness parent: /NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtcohe432/train/htmnpo_coherence

Map:   0%|          | 0/2250 [00:00<?, ? examples/s]
Map:  89%|████████▉ | 2000/2250 [00:00<00:00, 18068.62 examples/s]
Map: 100%|██████████| 2250/2250 [00:00<00:00, 17427.82 examples/s]

Map:   0%|          | 0/250 [00:00<?, ? examples/s]
Map: 100%|██████████| 250/250 [00:00<00:00, 13742.44 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, 190192.95 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 187681.40 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, 46775.93 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 45941.82 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, 303329.39 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 299621.68 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, 57393.32 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 56384.15 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_mbhtcohe433/train/htmnpo_coherence 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": 0.016, "correctness": 0.025, "coherence": -0.005, "conciseness": -0.004}, "P(pi>mu)": {"helpfulness": 0.508, "correctness": 0.512, "coherence": 0.497, "conciseness": 0.498}, "w_nbs": {"helpfulness": 0.0, "correctness": 0.0, "coherence": 0.5, "conciseness": 0.5}, "w_ks": {"helpfulness": 0.223, "correctness": 0.194, "coherence": 0.292, "conciseness": 0.292}, "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_mbhtcohe434] objs: helpfulness correctness coherence conciseness parent: /NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtcohe433/train/htmnpo_coherence

Map:   0%|          | 0/2250 [00:00<?, ? examples/s]
Map:  89%|████████▉ | 2000/2250 [00:00<00:00, 17435.33 examples/s]
Map: 100%|██████████| 2250/2250 [00:00<00:00, 16931.27 examples/s]

Map:   0%|          | 0/250 [00:00<?, ? examples/s]
Map: 100%|██████████| 250/250 [00:00<00:00, 12832.11 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, 208547.33 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 206819.72 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, 52043.68 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 51147.55 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, 292308.63 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 289226.88 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, 58455.57 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 57494.02 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-5 task=helpsteer2 anchor=/NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtcohe434/train/htmnpo_coherence 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": 0.01, "correctness": 0.018, "coherence": 0.009, "conciseness": 0.009}, "P(pi>mu)": {"helpfulness": 0.505, "correctness": 0.509, "coherence": 0.504, "conciseness": 0.504}, "w_nbs": {"helpfulness": 0.255, "correctness": 0.148, "coherence": 0.302, "conciseness": 0.295}, "w_ks": {"helpfulness": 0.254, "correctness": 0.225, "coherence": 0.261, "conciseness": 0.26}, "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_mbhtcohe435] objs: helpfulness correctness coherence conciseness parent: /NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtcohe434/train/htmnpo_coherence

Map:   0%|          | 0/2250 [00:00<?, ? examples/s]
Map:  89%|████████▉ | 2000/2250 [00:00<00:00, 17796.04 examples/s]
Map: 100%|██████████| 2250/2250 [00:00<00:00, 17224.19 examples/s]

Map:   0%|          | 0/250 [00:00<?, ? examples/s]
Map: 100%|██████████| 250/250 [00:00<00:00, 10879.04 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, 205782.47 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 204272.47 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, 52499.67 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 51707.48 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, 307310.04 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 304189.79 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, 59788.80 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 58779.98 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_mbhtcohe435/train/htmnpo_coherence 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": 0.02, "conciseness": -0.018}, "P(pi>mu)": {"helpfulness": 0.503, "correctness": 0.507, "coherence": 0.51, "conciseness": 0.491}, "w_nbs": {"helpfulness": 0.0, "correctness": 0.0, "coherence": 0.0, "conciseness": 0.999}, "w_ks": {"helpfulness": 0.262, "correctness": 0.235, "coherence": 0.21, "conciseness": 0.293}, "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_mbhtcohe436] objs: helpfulness correctness coherence conciseness parent: /NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtcohe435/train/htmnpo_coherence

Map:   0%|          | 0/2250 [00:00<?, ? examples/s]
Map:  57%|█████▋    | 1283/2250 [00:00<00:00, 10730.31 examples/s]
Map: 100%|██████████| 2250/2250 [00:00<00:00, 12282.64 examples/s]

Map:   0%|          | 0/250 [00:00<?, ? examples/s]
Map: 100%|██████████| 250/250 [00:00<00:00, 13556.43 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, 200872.35 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 199492.33 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, 53619.15 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 52798.39 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, 295586.31 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 292698.47 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, 57919.58 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 56956.87 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_mbhtcohe436/train/htmnpo_coherence 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": 0.013, "correctness": 0.016, "coherence": 0.002, "conciseness": -0.003}, "P(pi>mu)": {"helpfulness": 0.506, "correctness": 0.508, "coherence": 0.501, "conciseness": 0.498}, "w_nbs": {"helpfulness": 0.0, "correctness": 0.0, "coherence": 0.001, "conciseness": 0.999}, "w_ks": {"helpfulness": 0.228, "correctness": 0.217, "coherence": 0.273, "conciseness": 0.282}, "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_mbhtcohe437] objs: helpfulness correctness coherence conciseness parent: /NHNHOME/AIPR/sjkim/gtpref/helpsteer2/exp/rmpool_mbhtcohe436/train/htmnpo_coherence

Map:   0%|          | 0/2250 [00:00<?, ? examples/s]
Map:  89%|████████▉ | 2000/2250 [00:00<00:00, 17764.53 examples/s]
Map: 100%|██████████| 2250/2250 [00:00<00:00, 17402.60 examples/s]

Map:   0%|          | 0/250 [00:00<?, ? examples/s]
Map: 100%|██████████| 250/250 [00:00<00:00, 13839.30 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, 207465.35 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 205975.60 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, 50108.76 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 49379.61 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, 300585.55 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 2250/2250 [00:00<00:00, 297590.31 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, 61169.99 examples/s]
Saving the dataset (1/1 shards): 100%|██████████| 250/250 [00:00<00:00, 60107.54 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
{
  "label": "qwen32_helpsteer2_htmnpo_coherence",
  "n_objectives": 4,
  "surplus": {
    "helpfulness": 0.1106,
    "correctness": 0.0934,
    "coherence": 0.0854,
    "conciseness": 0.0691
  },
  "hoeffding_margin": {
    "helpfulness": 0.0357,
    "correctness": 0.0359,
    "coherence": 0.0356,
    "conciseness": 0.0356
  },
  "worst_surplus": 0.0691,
  "individual_rationality": true,
  "nash_welfare_logsum": -9.7059
}
EVAL_helpsteer2_DONE aiprlab-polymer
MB_STAGES_DONE