Upload folder using huggingface_hub
Browse files- README.md +19 -9
- decider/engine.py +1 -2
- decider/infer.py +9 -1
- decider_config.json +1 -0
- eval_results.json +840 -761
- model.safetensors +1 -1
- tokenizer_config.json +1 -1
README.md
CHANGED
|
@@ -17,7 +17,9 @@ schema violations. It is meant to be called from software, not chatted with.
|
|
| 17 |
Base model: [Qwen/Qwen3.5-2B-Base](https://huggingface.co/Qwen/Qwen3.5-2B-Base) (1.9B parameters),
|
| 18 |
fully fine-tuned for one epoch (942k examples, 183M tokens, 2.5 hours on one
|
| 19 |
NVIDIA GH200) with cross-entropy, a proper scoring rule, on a mixture of 64
|
| 20 |
-
public decision datasets
|
|
|
|
|
|
|
| 21 |
|
| 22 |
## Usage
|
| 23 |
|
|
@@ -76,13 +78,16 @@ the supplied candidates rather than a fixed head.
|
|
| 76 |
|
| 77 |
## Training data
|
| 78 |
|
| 79 |
-
|
| 80 |
intent detection, ticket routing, topic classification, sentiment, emotion,
|
| 81 |
moderation (toxicity, hate, spam, jailbreak, safety), NLI, paraphrase, fact
|
| 82 |
verification, passage relevance, reading comprehension, multiple-choice QA,
|
| 83 |
ordinal rating scales (HelpSteer2 attributes, STS-B, hate-speech intensity,
|
| 84 |
LIAR2 truthfulness), pairwise response preference (HelpSteer3, UltraFeedback,
|
| 85 |
SHP, HH-RLHF) and tool selection (Glaive, ToolACE).
|
|
|
|
|
|
|
|
|
|
| 86 |
Abstention augmentation: in 10% of questions with three or more options the
|
| 87 |
gold option is removed and "none of the above" becomes the answer.
|
| 88 |
|
|
@@ -94,8 +99,8 @@ gold option is removed and "none of the above" becomes the answer.
|
|
| 94 |
| Qwen3.5-2B-Base, zero-shot | held-out (23) | 0.642 | 0.853 | 0.460 | 0.105 | 0.242 | 0.685 |
|
| 95 |
| Qwen3.5-4B-Base, zero-shot | in-task (64) | 0.695 | 0.768 | 0.405 | 0.090 | 0.206 | 0.742 |
|
| 96 |
| Qwen3.5-4B-Base, zero-shot | held-out (23) | 0.711 | 0.734 | 0.390 | 0.089 | 0.169 | 0.761 |
|
| 97 |
-
| **this model** | in-task (
|
| 98 |
-
| **this model** | held-out (23) | 0.
|
| 99 |
|
| 100 |
|
| 101 |
Per-task accuracy / ECE on the held-out datasets:
|
|
@@ -191,7 +196,8 @@ Per-task accuracy / ECE on the held-out datasets:
|
|
| 191 |
| tweet_irony | 0.511 / 0.158 | 0.676 / 0.073 | 0.769 / 0.048 |
|
| 192 |
|
| 193 |
|
| 194 |
-
|
|
|
|
| 195 |
seen in training: TREC, BBC news, PAWS, SciQ, Social IQa, StrategyQA, PubMedQA,
|
| 196 |
TruthfulQA, tweet irony, financial sentiment, ADE, MASSIVE scenario, student
|
| 197 |
question categories, Dolly categories, CR reviews, Financial PhraseBank,
|
|
@@ -212,10 +218,14 @@ questions each:
|
|
| 212 |
| setting | p50 latency | throughput |
|
| 213 |
|---|---|---|
|
| 214 |
| single request, eager PyTorch | 49 ms | |
|
| 215 |
-
| single request, CUDA
|
| 216 |
-
| batch of 32, in-process |
|
| 217 |
-
|
|
| 218 |
-
| HTTP server,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 219 |
|
| 220 |
## Limitations
|
| 221 |
|
|
|
|
| 17 |
Base model: [Qwen/Qwen3.5-2B-Base](https://huggingface.co/Qwen/Qwen3.5-2B-Base) (1.9B parameters),
|
| 18 |
fully fine-tuned for one epoch (942k examples, 183M tokens, 2.5 hours on one
|
| 19 |
NVIDIA GH200) with cross-entropy, a proper scoring rule, on a mixture of 64
|
| 20 |
+
public decision datasets, then continued for one epoch on 45k situation-to-action
|
| 21 |
+
examples (agent trajectories, web element choice, synthetic situations, game states)
|
| 22 |
+
with a replay of the general mixture (v4).
|
| 23 |
|
| 24 |
## Usage
|
| 25 |
|
|
|
|
| 78 |
|
| 79 |
## Training data
|
| 80 |
|
| 81 |
+
69 public datasets plus synthetic situations, up to 20k examples each (`decider/data.py`, `decider/data2.py`, `decider/data3.py`):
|
| 82 |
intent detection, ticket routing, topic classification, sentiment, emotion,
|
| 83 |
moderation (toxicity, hate, spam, jailbreak, safety), NLI, paraphrase, fact
|
| 84 |
verification, passage relevance, reading comprehension, multiple-choice QA,
|
| 85 |
ordinal rating scales (HelpSteer2 attributes, STS-B, hate-speech intensity,
|
| 86 |
LIAR2 truthfulness), pairwise response preference (HelpSteer3, UltraFeedback,
|
| 87 |
SHP, HH-RLHF) and tool selection (Glaive, ToolACE).
|
| 88 |
+
v4 adds next-action choice from agent trajectories (AgentGym AgentTraj-L), web element
|
| 89 |
+
choice (Mind2Web), 1.5k synthetic situations written by Qwen3.5-27B, and teacher-labelled
|
| 90 |
+
states from Pong, Breakout, CliffWalking, MiniGrid and Super Mario Bros.
|
| 91 |
Abstention augmentation: in 10% of questions with three or more options the
|
| 92 |
gold option is removed and "none of the above" becomes the answer.
|
| 93 |
|
|
|
|
| 99 |
| Qwen3.5-2B-Base, zero-shot | held-out (23) | 0.642 | 0.853 | 0.460 | 0.105 | 0.242 | 0.685 |
|
| 100 |
| Qwen3.5-4B-Base, zero-shot | in-task (64) | 0.695 | 0.768 | 0.405 | 0.090 | 0.206 | 0.742 |
|
| 101 |
| Qwen3.5-4B-Base, zero-shot | held-out (23) | 0.711 | 0.734 | 0.390 | 0.089 | 0.169 | 0.761 |
|
| 102 |
+
| **this model (v4)** | in-task (69) | 0.813 | 0.450 | 0.250 | 0.033 | 0.093 | 0.864 |
|
| 103 |
+
| **this model (v4)** | held-out (23) | 0.748 | 0.634 | 0.345 | 0.076 | 0.137 | 0.803 |
|
| 104 |
|
| 105 |
|
| 106 |
Per-task accuracy / ECE on the held-out datasets:
|
|
|
|
| 196 |
| tweet_irony | 0.511 / 0.158 | 0.676 / 0.073 | 0.769 / 0.048 |
|
| 197 |
|
| 198 |
|
| 199 |
+
Probabilities use a temperature of 1.15 fitted on in-task data (stored in `decider_config.json`, applied by the helper).
|
| 200 |
+
*In-task* = test splits of the training datasets (in-task rows for the zero-shot baselines cover the original 64). *Held-out* = 23 datasets never
|
| 201 |
seen in training: TREC, BBC news, PAWS, SciQ, Social IQa, StrategyQA, PubMedQA,
|
| 202 |
TruthfulQA, tweet irony, financial sentiment, ADE, MASSIVE scenario, student
|
| 203 |
question categories, Dolly categories, CR reviews, Financial PhraseBank,
|
|
|
|
| 218 |
| setting | p50 latency | throughput |
|
| 219 |
|---|---|---|
|
| 220 |
| single request, eager PyTorch | 49 ms | |
|
| 221 |
+
| single request, CUDA graphs + torch.compile (helper default) | 4.0 ms | |
|
| 222 |
+
| batch of 32, in-process, bf16 | 70 ms | ~1370 decisions/s |
|
| 223 |
+
| batch of 32, in-process, FP8 linears | 58 ms | ~1670 decisions/s |
|
| 224 |
+
| HTTP server (FP8), 1 client | 6.8 ms | 134 req/s |
|
| 225 |
+
| HTTP server (FP8), 64 clients | 126 ms | 431 req/s, 2152 decisions/s |
|
| 226 |
+
|
| 227 |
+
FP8 (e4m3 weights, per-token activation scales) changes accuracy and calibration by
|
| 228 |
+
less than the evaluation noise (18-task check: accuracy 0.833 vs 0.835, ECE equal).
|
| 229 |
|
| 230 |
## Limitations
|
| 231 |
|
decider/engine.py
CHANGED
|
@@ -4,7 +4,7 @@ Right padding + causal layers => pad positions never influence earlier slots, so
|
|
| 4 |
mask is needed and every (B, T) bucket can be captured once and replayed. The graph outputs
|
| 5 |
option-letter logits for all positions [B, T, K]; slots are gathered outside.
|
| 6 |
"""
|
| 7 |
-
import time, torch, torch.nn.functional as F
|
| 8 |
from .model import DecisionModel, collate
|
| 9 |
from .prompt import build, MAX_OPTIONS
|
| 10 |
|
|
@@ -58,7 +58,6 @@ class Engine:
|
|
| 58 |
from .fp8 import convert_to_fp8
|
| 59 |
self.cfg["fp8_layers"] = convert_to_fp8(self.core)
|
| 60 |
if compile:
|
| 61 |
-
import torch._dynamo
|
| 62 |
torch._dynamo.config.cache_size_limit = 128
|
| 63 |
self._fwd_impl = torch.compile(self._fwd_eager, dynamic=False)
|
| 64 |
else:
|
|
|
|
| 4 |
mask is needed and every (B, T) bucket can be captured once and replayed. The graph outputs
|
| 5 |
option-letter logits for all positions [B, T, K]; slots are gathered outside.
|
| 6 |
"""
|
| 7 |
+
import time, torch, torch._dynamo, torch.nn.functional as F
|
| 8 |
from .model import DecisionModel, collate
|
| 9 |
from .prompt import build, MAX_OPTIONS
|
| 10 |
|
|
|
|
| 58 |
from .fp8 import convert_to_fp8
|
| 59 |
self.cfg["fp8_layers"] = convert_to_fp8(self.core)
|
| 60 |
if compile:
|
|
|
|
| 61 |
torch._dynamo.config.cache_size_limit = 128
|
| 62 |
self._fwd_impl = torch.compile(self._fwd_eager, dynamic=False)
|
| 63 |
else:
|
decider/infer.py
CHANGED
|
@@ -26,7 +26,15 @@ class Example:
|
|
| 26 |
class Decider:
|
| 27 |
"""use_graphs=True (default on CUDA) routes scoring through decider.engine.Engine: shape-bucketed
|
| 28 |
CUDA graphs, ~7x lower single-request latency than eager. Set False for CPU or debugging."""
|
| 29 |
-
def __init__(self, path, device="cuda", dtype=torch.bfloat16, temperature=
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 30 |
if use_graphs is None:
|
| 31 |
use_graphs = str(device).startswith("cuda")
|
| 32 |
if use_graphs:
|
|
|
|
| 26 |
class Decider:
|
| 27 |
"""use_graphs=True (default on CUDA) routes scoring through decider.engine.Engine: shape-bucketed
|
| 28 |
CUDA graphs, ~7x lower single-request latency than eager. Set False for CPU or debugging."""
|
| 29 |
+
def __init__(self, path, device="cuda", dtype=torch.bfloat16, temperature=None, abstain_below=0.0, use_graphs=None):
|
| 30 |
+
if temperature is None: # model folder may carry a fitted temperature (decider_config.json)
|
| 31 |
+
import json, os
|
| 32 |
+
try:
|
| 33 |
+
from huggingface_hub import hf_hub_download
|
| 34 |
+
cfg_path = os.path.join(path, "decider_config.json") if os.path.isdir(path) else hf_hub_download(path, "decider_config.json")
|
| 35 |
+
temperature = float(json.load(open(cfg_path)).get("temperature", 1.0))
|
| 36 |
+
except Exception:
|
| 37 |
+
temperature = 1.0
|
| 38 |
if use_graphs is None:
|
| 39 |
use_graphs = str(device).startswith("cuda")
|
| 40 |
if use_graphs:
|
decider_config.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"temperature": 1.15, "version": "v4", "base": "Qwen/Qwen3.5-2B-Base"}
|
eval_results.json
CHANGED
|
@@ -2,1282 +2,1361 @@
|
|
| 2 |
"results": {
|
| 3 |
"clinc_oos": {
|
| 4 |
"n": 1500,
|
| 5 |
-
"acc": 0.
|
| 6 |
-
"nll": 0.
|
| 7 |
-
"brier": 0.
|
| 8 |
-
"ece": 0.
|
| 9 |
-
"aurc": 0.
|
| 10 |
"acc_at_80": 0.9991666666666666,
|
| 11 |
"acc_at_50": 1.0,
|
| 12 |
"chance": 0.1,
|
| 13 |
-
"mean_conf": 0.
|
| 14 |
-
"sec":
|
| 15 |
"heldout": false
|
| 16 |
},
|
| 17 |
"massive_intent": {
|
| 18 |
"n": 1500,
|
| 19 |
-
"acc": 0.
|
| 20 |
-
"nll": 0.
|
| 21 |
-
"brier": 0.
|
| 22 |
-
"ece": 0.
|
| 23 |
-
"aurc": 0.
|
| 24 |
"acc_at_80": 0.9933333333333333,
|
| 25 |
-
"acc_at_50": 0.
|
| 26 |
"chance": 0.1,
|
| 27 |
-
"mean_conf": 0.
|
| 28 |
-
"sec":
|
| 29 |
"heldout": false
|
| 30 |
},
|
| 31 |
"massive_scenario": {
|
| 32 |
"n": 1500,
|
| 33 |
-
"acc": 0.
|
| 34 |
-
"nll": 0.
|
| 35 |
-
"brier": 0.
|
| 36 |
-
"ece": 0.
|
| 37 |
-
"aurc": 0.
|
| 38 |
-
"acc_at_80": 0.
|
| 39 |
-
"acc_at_50": 0.
|
| 40 |
"chance": 0.1,
|
| 41 |
-
"mean_conf": 0.
|
| 42 |
-
"sec":
|
| 43 |
"heldout": true
|
| 44 |
},
|
| 45 |
"bitext_support": {
|
| 46 |
"n": 3000,
|
| 47 |
-
"acc": 0.
|
| 48 |
-
"nll": 0.
|
| 49 |
-
"brier": 0.
|
| 50 |
-
"ece": 0.
|
| 51 |
-
"aurc": 1.
|
| 52 |
"acc_at_80": 1.0,
|
| 53 |
"acc_at_50": 1.0,
|
| 54 |
"chance": 0.09999999999999998,
|
| 55 |
-
"mean_conf": 0.
|
| 56 |
-
"sec":
|
| 57 |
"heldout": false,
|
| 58 |
"per_q": [
|
| 59 |
-
0.
|
| 60 |
-
0.
|
| 61 |
]
|
| 62 |
},
|
| 63 |
"support_tickets": {
|
| 64 |
"n": 4500,
|
| 65 |
-
"acc": 0.
|
| 66 |
-
"nll": 0.
|
| 67 |
-
"brier": 0.
|
| 68 |
-
"ece": 0.
|
| 69 |
-
"aurc": 0.
|
| 70 |
-
"acc_at_80": 0.
|
| 71 |
-
"acc_at_50": 0.
|
| 72 |
"chance": 0.22777777777777777,
|
| 73 |
-
"mean_conf": 0.
|
| 74 |
-
"sec":
|
| 75 |
"heldout": false,
|
| 76 |
"per_q": [
|
| 77 |
-
0.
|
| 78 |
-
0.
|
| 79 |
-
0.
|
| 80 |
]
|
| 81 |
},
|
| 82 |
"ag_news": {
|
| 83 |
"n": 1500,
|
| 84 |
"acc": 0.9173333333333333,
|
| 85 |
-
"nll": 0.
|
| 86 |
-
"brier": 0.
|
| 87 |
-
"ece": 0.
|
| 88 |
-
"aurc": 0.
|
| 89 |
"acc_at_80": 0.9716666666666667,
|
| 90 |
-
"acc_at_50": 0.
|
| 91 |
"chance": 0.25,
|
| 92 |
-
"mean_conf": 0.
|
| 93 |
-
"sec":
|
| 94 |
"heldout": false
|
| 95 |
},
|
| 96 |
"dbpedia": {
|
| 97 |
"n": 1500,
|
| 98 |
"acc": 0.9913333333333333,
|
| 99 |
-
"nll": 0.
|
| 100 |
-
"brier": 0.
|
| 101 |
-
"ece": 0.
|
| 102 |
-
"aurc": 0.
|
| 103 |
"acc_at_80": 1.0,
|
| 104 |
"acc_at_50": 1.0,
|
| 105 |
"chance": 0.1,
|
| 106 |
-
"mean_conf": 0.
|
| 107 |
-
"sec":
|
| 108 |
"heldout": false
|
| 109 |
},
|
| 110 |
"yahoo_topics": {
|
| 111 |
"n": 1500,
|
| 112 |
-
"acc": 0.
|
| 113 |
-
"nll": 0.
|
| 114 |
-
"brier": 0.
|
| 115 |
-
"ece": 0.
|
| 116 |
-
"aurc": 0.
|
| 117 |
-
"acc_at_80": 0.
|
| 118 |
-
"acc_at_50": 0.
|
| 119 |
"chance": 0.1,
|
| 120 |
-
"mean_conf": 0.
|
| 121 |
-
"sec":
|
| 122 |
"heldout": false
|
| 123 |
},
|
| 124 |
"newsgroups": {
|
| 125 |
"n": 1462,
|
| 126 |
-
"acc": 0.
|
| 127 |
-
"nll": 0.
|
| 128 |
-
"brier": 0.
|
| 129 |
-
"ece": 0.
|
| 130 |
-
"aurc": 0.
|
| 131 |
"acc_at_80": 0.917094017094017,
|
| 132 |
"acc_at_50": 0.9835841313269493,
|
| 133 |
"chance": 0.09999999999999999,
|
| 134 |
-
"mean_conf": 0.
|
| 135 |
-
"sec":
|
| 136 |
"heldout": false
|
| 137 |
},
|
| 138 |
"bbc_news": {
|
| 139 |
"n": 1000,
|
| 140 |
-
"acc": 0.
|
| 141 |
-
"nll": 0.
|
| 142 |
-
"brier": 0.
|
| 143 |
-
"ece": 0.
|
| 144 |
-
"aurc": 0.
|
| 145 |
-
"acc_at_80": 0.
|
| 146 |
"acc_at_50": 1.0,
|
| 147 |
"chance": 0.20000000000000004,
|
| 148 |
-
"mean_conf": 0.
|
| 149 |
-
"sec": 5.
|
| 150 |
"heldout": true
|
| 151 |
},
|
| 152 |
"trec": {
|
| 153 |
"n": 500,
|
| 154 |
-
"acc": 0.
|
| 155 |
-
"nll": 0.
|
| 156 |
-
"brier": 0.
|
| 157 |
-
"ece": 0.
|
| 158 |
-
"aurc": 0.
|
| 159 |
-
"acc_at_80": 0.
|
| 160 |
-
"acc_at_50": 0.
|
| 161 |
"chance": 0.16666666666666663,
|
| 162 |
-
"mean_conf": 0.
|
| 163 |
-
"sec":
|
| 164 |
"heldout": true
|
| 165 |
},
|
| 166 |
"student_questions": {
|
| 167 |
"n": 1500,
|
| 168 |
-
"acc": 0.
|
| 169 |
-
"nll": 0.
|
| 170 |
-
"brier": 0.
|
| 171 |
-
"ece": 0.
|
| 172 |
-
"aurc": 0.
|
| 173 |
-
"acc_at_80": 0.
|
| 174 |
"acc_at_50": 0.992,
|
| 175 |
"chance": 0.25,
|
| 176 |
-
"mean_conf": 0.
|
| 177 |
-
"sec": 3.
|
| 178 |
"heldout": true
|
| 179 |
},
|
| 180 |
"dolly_category": {
|
| 181 |
"n": 1500,
|
| 182 |
-
"acc": 0.
|
| 183 |
-
"nll": 2.
|
| 184 |
-
"brier": 0.
|
| 185 |
-
"ece": 0.
|
| 186 |
-
"aurc": 0.
|
| 187 |
-
"acc_at_80": 0.
|
| 188 |
-
"acc_at_50": 0.
|
| 189 |
"chance": 0.125,
|
| 190 |
-
"mean_conf": 0.
|
| 191 |
-
"sec": 3.
|
| 192 |
"heldout": true
|
| 193 |
},
|
| 194 |
"imdb": {
|
| 195 |
"n": 1500,
|
| 196 |
-
"acc": 0.
|
| 197 |
-
"nll": 0.
|
| 198 |
-
"brier": 0.
|
| 199 |
-
"ece": 0.
|
| 200 |
-
"aurc": 0.
|
| 201 |
"acc_at_80": 0.9983333333333333,
|
| 202 |
"acc_at_50": 1.0,
|
| 203 |
"chance": 0.5,
|
| 204 |
-
"mean_conf": 0.
|
| 205 |
-
"sec": 5.
|
| 206 |
"heldout": false
|
| 207 |
},
|
| 208 |
"sst2": {
|
| 209 |
"n": 872,
|
| 210 |
-
"acc": 0.
|
| 211 |
-
"nll": 0.
|
| 212 |
-
"brier": 0.
|
| 213 |
-
"ece": 0.
|
| 214 |
-
"aurc": 0.
|
| 215 |
"acc_at_80": 0.9899713467048711,
|
| 216 |
"acc_at_50": 0.9931192660550459,
|
| 217 |
"chance": 0.5,
|
| 218 |
-
"mean_conf": 0.
|
| 219 |
-
"sec":
|
| 220 |
"heldout": false
|
| 221 |
},
|
| 222 |
"sst5": {
|
| 223 |
"n": 1500,
|
| 224 |
-
"acc": 0.
|
| 225 |
-
"nll": 0.
|
| 226 |
-
"brier": 0.
|
| 227 |
-
"ece": 0.
|
| 228 |
-
"aurc": 0.
|
| 229 |
"acc_at_80": 0.6466666666666666,
|
| 230 |
-
"acc_at_50": 0.
|
| 231 |
"chance": 0.2,
|
| 232 |
-
"mean_conf": 0.
|
| 233 |
-
"sec": 2.
|
| 234 |
"heldout": false
|
| 235 |
},
|
| 236 |
"yelp": {
|
| 237 |
"n": 1500,
|
| 238 |
-
"acc": 0.
|
| 239 |
-
"nll": 0.
|
| 240 |
-
"brier": 0.
|
| 241 |
-
"ece": 0.
|
| 242 |
-
"aurc": 0.
|
| 243 |
-
"acc_at_80": 0.
|
| 244 |
"acc_at_50": 0.8146666666666667,
|
| 245 |
"chance": 0.2,
|
| 246 |
-
"mean_conf": 0.
|
| 247 |
-
"sec": 4.
|
| 248 |
"heldout": false
|
| 249 |
},
|
| 250 |
"amazon_stars": {
|
| 251 |
"n": 1500,
|
| 252 |
-
"acc": 0.
|
| 253 |
-
"nll": 0.
|
| 254 |
-
"brier": 0.
|
| 255 |
-
"ece": 0.
|
| 256 |
-
"aurc": 0.
|
| 257 |
-
"acc_at_80": 0.
|
| 258 |
-
"acc_at_50": 0.
|
| 259 |
"chance": 0.2,
|
| 260 |
-
"mean_conf": 0.
|
| 261 |
-
"sec":
|
| 262 |
"heldout": false
|
| 263 |
},
|
| 264 |
"emotion": {
|
| 265 |
"n": 1500,
|
| 266 |
-
"acc": 0.
|
| 267 |
-
"nll": 0.
|
| 268 |
-
"brier": 0.
|
| 269 |
-
"ece": 0.
|
| 270 |
-
"aurc": 0.
|
| 271 |
-
"acc_at_80": 0.
|
| 272 |
"acc_at_50": 0.9813333333333333,
|
| 273 |
"chance": 0.16666666666666666,
|
| 274 |
-
"mean_conf": 0.
|
| 275 |
-
"sec": 2.
|
| 276 |
"heldout": false
|
| 277 |
},
|
| 278 |
"go_emotions": {
|
| 279 |
"n": 1500,
|
| 280 |
-
"acc": 0.
|
| 281 |
-
"nll": 0.
|
| 282 |
-
"brier": 0.
|
| 283 |
-
"ece": 0.
|
| 284 |
-
"aurc": 0.
|
| 285 |
-
"acc_at_80": 0.
|
| 286 |
-
"acc_at_50": 0.
|
| 287 |
"chance": 0.1,
|
| 288 |
-
"mean_conf": 0.
|
| 289 |
-
"sec":
|
| 290 |
"heldout": false
|
| 291 |
},
|
| 292 |
"tweet_sentiment": {
|
| 293 |
"n": 1500,
|
| 294 |
-
"acc": 0.
|
| 295 |
-
"nll": 0.
|
| 296 |
-
"brier": 0.
|
| 297 |
-
"ece": 0.
|
| 298 |
-
"aurc": 0.
|
| 299 |
-
"acc_at_80": 0.
|
| 300 |
"acc_at_50": 0.868,
|
| 301 |
"chance": 0.3333333333333333,
|
| 302 |
-
"mean_conf": 0.
|
| 303 |
"sec": 2.4,
|
| 304 |
"heldout": false
|
| 305 |
},
|
| 306 |
"tweet_emotion": {
|
| 307 |
"n": 1421,
|
| 308 |
-
"acc": 0.
|
| 309 |
-
"nll": 0.
|
| 310 |
-
"brier": 0.
|
| 311 |
-
"ece": 0.
|
| 312 |
-
"aurc": 0.
|
| 313 |
-
"acc_at_80": 0.
|
| 314 |
-
"acc_at_50": 0.
|
| 315 |
"chance": 0.25,
|
| 316 |
-
"mean_conf": 0.
|
| 317 |
-
"sec":
|
| 318 |
"heldout": false
|
| 319 |
},
|
| 320 |
"tweet_irony": {
|
| 321 |
"n": 784,
|
| 322 |
-
"acc": 0.
|
| 323 |
-
"nll": 0.
|
| 324 |
-
"brier": 0.
|
| 325 |
-
"ece": 0.
|
| 326 |
-
"aurc": 0.
|
| 327 |
-
"acc_at_80": 0.
|
| 328 |
-
"acc_at_50": 0.
|
| 329 |
"chance": 0.5,
|
| 330 |
-
"mean_conf": 0.
|
| 331 |
-
"sec": 1.
|
| 332 |
"heldout": true
|
| 333 |
},
|
| 334 |
"fin_sentiment": {
|
| 335 |
"n": 1500,
|
| 336 |
-
"acc": 0.
|
| 337 |
-
"nll": 0.
|
| 338 |
-
"brier": 0.
|
| 339 |
-
"ece": 0.
|
| 340 |
-
"aurc": 0.
|
| 341 |
-
"acc_at_80": 0.
|
| 342 |
-
"acc_at_50": 0.
|
| 343 |
"chance": 0.3333333333333333,
|
| 344 |
-
"mean_conf": 0.
|
| 345 |
-
"sec": 2.
|
| 346 |
"heldout": true
|
| 347 |
},
|
| 348 |
"cr_reviews": {
|
| 349 |
"n": 753,
|
| 350 |
-
"acc": 0.
|
| 351 |
-
"nll": 0.
|
| 352 |
-
"brier": 0.
|
| 353 |
-
"ece": 0.
|
| 354 |
-
"aurc": 0.
|
| 355 |
-
"acc_at_80": 0.
|
| 356 |
-
"acc_at_50": 0.
|
| 357 |
"chance": 0.5,
|
| 358 |
-
"mean_conf": 0.
|
| 359 |
-
"sec": 1.
|
| 360 |
"heldout": true
|
| 361 |
},
|
| 362 |
"counterfactual": {
|
| 363 |
"n": 1500,
|
| 364 |
-
"acc": 0.
|
| 365 |
-
"nll": 0.
|
| 366 |
-
"brier": 0.
|
| 367 |
-
"ece": 0.
|
| 368 |
-
"aurc": 0.
|
| 369 |
-
"acc_at_80": 0.
|
| 370 |
-
"acc_at_50":
|
| 371 |
"chance": 0.5,
|
| 372 |
-
"mean_conf": 0.
|
| 373 |
-
"sec": 2.
|
| 374 |
"heldout": false
|
| 375 |
},
|
| 376 |
"subjectivity": {
|
| 377 |
"n": 1500,
|
| 378 |
-
"acc": 0.
|
| 379 |
-
"nll": 0.
|
| 380 |
-
"brier": 0.
|
| 381 |
-
"ece": 0.
|
| 382 |
-
"aurc": 0.
|
| 383 |
"acc_at_80": 0.9916666666666667,
|
| 384 |
"acc_at_50": 0.9973333333333333,
|
| 385 |
"chance": 0.5,
|
| 386 |
-
"mean_conf": 0.
|
| 387 |
"sec": 2.4,
|
| 388 |
"heldout": false
|
| 389 |
},
|
| 390 |
"tweet_offensive": {
|
| 391 |
"n": 860,
|
| 392 |
-
"acc": 0.
|
| 393 |
-
"nll": 0.
|
| 394 |
-
"brier": 0.
|
| 395 |
-
"ece": 0.
|
| 396 |
-
"aurc": 0.
|
| 397 |
-
"acc_at_80": 0.
|
| 398 |
-
"acc_at_50": 0.
|
| 399 |
"chance": 0.5,
|
| 400 |
-
"mean_conf": 0.
|
| 401 |
-
"sec": 1.
|
| 402 |
"heldout": false
|
| 403 |
},
|
| 404 |
"tweet_hate": {
|
| 405 |
"n": 1500,
|
| 406 |
-
"acc": 0.
|
| 407 |
-
"nll": 0.
|
| 408 |
-
"brier": 0.
|
| 409 |
-
"ece": 0.
|
| 410 |
-
"aurc": 0.
|
| 411 |
"acc_at_80": 0.535,
|
| 412 |
-
"acc_at_50": 0.
|
| 413 |
"chance": 0.5,
|
| 414 |
-
"mean_conf": 0.
|
| 415 |
-
"sec": 2.
|
| 416 |
"heldout": false
|
| 417 |
},
|
| 418 |
"hate_offensive": {
|
| 419 |
"n": 1500,
|
| 420 |
-
"acc": 0.
|
| 421 |
-
"nll": 0.
|
| 422 |
-
"brier": 0.
|
| 423 |
-
"ece": 0.
|
| 424 |
-
"aurc": 0.
|
| 425 |
"acc_at_80": 0.9716666666666667,
|
| 426 |
-
"acc_at_50": 0.
|
| 427 |
"chance": 0.3333333333333333,
|
| 428 |
-
"mean_conf": 0.
|
| 429 |
-
"sec": 2.
|
| 430 |
"heldout": false
|
| 431 |
},
|
| 432 |
"civil_comments": {
|
| 433 |
"n": 7500,
|
| 434 |
-
"acc": 0.
|
| 435 |
-
"nll": 0.
|
| 436 |
-
"brier": 0.
|
| 437 |
-
"ece": 0.
|
| 438 |
-
"aurc": 0.
|
| 439 |
-
"acc_at_80": 0.
|
| 440 |
"acc_at_50": 0.9997333333333334,
|
| 441 |
"chance": 0.5,
|
| 442 |
-
"mean_conf": 0.
|
| 443 |
-
"sec": 4.
|
| 444 |
"heldout": false,
|
| 445 |
"per_q": [
|
| 446 |
-
0.
|
| 447 |
-
0.
|
| 448 |
-
0.
|
| 449 |
-
0.
|
| 450 |
0.9693333333333334
|
| 451 |
]
|
| 452 |
},
|
| 453 |
"toxic_chat": {
|
| 454 |
"n": 3000,
|
| 455 |
-
"acc": 0.
|
| 456 |
-
"nll": 0.
|
| 457 |
-
"brier": 0.
|
| 458 |
-
"ece": 0.
|
| 459 |
-
"aurc": 0.
|
| 460 |
-
"acc_at_80": 0.
|
| 461 |
-
"acc_at_50":
|
| 462 |
"chance": 0.5,
|
| 463 |
-
"mean_conf": 0.
|
| 464 |
-
"sec":
|
| 465 |
"heldout": false,
|
| 466 |
"per_q": [
|
| 467 |
-
0.
|
| 468 |
-
0.
|
| 469 |
]
|
| 470 |
},
|
| 471 |
"sms_spam": {
|
| 472 |
"n": 1000,
|
| 473 |
-
"acc": 0.
|
| 474 |
-
"nll": 0.
|
| 475 |
-
"brier": 0.
|
| 476 |
-
"ece": 0.
|
| 477 |
-
"aurc": 0.
|
| 478 |
"acc_at_80": 0.99625,
|
| 479 |
"acc_at_50": 0.998,
|
| 480 |
"chance": 0.5,
|
| 481 |
-
"mean_conf": 0.
|
| 482 |
-
"sec":
|
| 483 |
"heldout": false
|
| 484 |
},
|
| 485 |
"enron_spam": {
|
| 486 |
"n": 1500,
|
| 487 |
-
"acc": 0.
|
| 488 |
-
"nll": 0.
|
| 489 |
-
"brier": 0.
|
| 490 |
-
"ece": 0.
|
| 491 |
-
"aurc": 0.
|
| 492 |
"acc_at_80": 1.0,
|
| 493 |
"acc_at_50": 1.0,
|
| 494 |
"chance": 0.5,
|
| 495 |
-
"mean_conf": 0.
|
| 496 |
-
"sec":
|
| 497 |
"heldout": false
|
| 498 |
},
|
| 499 |
"insincere_questions": {
|
| 500 |
"n": 1500,
|
| 501 |
"acc": 0.9553333333333334,
|
| 502 |
-
"nll": 0.
|
| 503 |
-
"brier": 0.
|
| 504 |
-
"ece": 0.
|
| 505 |
-
"aurc": 0.
|
| 506 |
"acc_at_80": 0.9975,
|
| 507 |
"acc_at_50": 0.9986666666666667,
|
| 508 |
"chance": 0.5,
|
| 509 |
-
"mean_conf": 0.
|
| 510 |
"sec": 2.4,
|
| 511 |
"heldout": false
|
| 512 |
},
|
| 513 |
"ade": {
|
| 514 |
"n": 1500,
|
| 515 |
-
"acc": 0.
|
| 516 |
-
"nll": 0.
|
| 517 |
-
"brier": 0.
|
| 518 |
-
"ece": 0.
|
| 519 |
-
"aurc": 0.
|
| 520 |
-
"acc_at_80": 0.
|
| 521 |
-
"acc_at_50": 0.
|
| 522 |
"chance": 0.5,
|
| 523 |
-
"mean_conf": 0.
|
| 524 |
-
"sec": 2.
|
| 525 |
"heldout": true
|
| 526 |
},
|
| 527 |
"snli": {
|
| 528 |
"n": 1476,
|
| 529 |
-
"acc": 0.
|
| 530 |
-
"nll": 0.
|
| 531 |
-
"brier": 0.
|
| 532 |
-
"ece": 0.
|
| 533 |
-
"aurc": 0.
|
| 534 |
"acc_at_80": 0.9729043183742591,
|
| 535 |
-
"acc_at_50": 0.
|
| 536 |
"chance": 0.3333333333333333,
|
| 537 |
-
"mean_conf": 0.
|
| 538 |
-
"sec":
|
| 539 |
"heldout": false
|
| 540 |
},
|
| 541 |
"mnli": {
|
| 542 |
"n": 1500,
|
| 543 |
-
"acc": 0.
|
| 544 |
-
"nll": 0.
|
| 545 |
-
"brier": 0.
|
| 546 |
-
"ece": 0.
|
| 547 |
-
"aurc": 0.
|
| 548 |
-
"acc_at_80": 0.
|
| 549 |
-
"acc_at_50": 0.
|
| 550 |
"chance": 0.3333333333333333,
|
| 551 |
-
"mean_conf": 0.
|
| 552 |
-
"sec":
|
| 553 |
"heldout": false
|
| 554 |
},
|
| 555 |
"rte": {
|
| 556 |
"n": 277,
|
| 557 |
-
"acc": 0.
|
| 558 |
-
"nll": 0.
|
| 559 |
-
"brier": 0.
|
| 560 |
-
"ece": 0.
|
| 561 |
-
"aurc": 0.
|
| 562 |
-
"acc_at_80": 0.
|
| 563 |
-
"acc_at_50": 0.
|
| 564 |
"chance": 0.5,
|
| 565 |
-
"mean_conf": 0.
|
| 566 |
-
"sec": 0.
|
| 567 |
"heldout": false
|
| 568 |
},
|
| 569 |
"qnli": {
|
| 570 |
"n": 1500,
|
| 571 |
-
"acc": 0.
|
| 572 |
-
"nll": 0.
|
| 573 |
-
"brier": 0.
|
| 574 |
-
"ece": 0.
|
| 575 |
-
"aurc": 0.
|
| 576 |
"acc_at_80": 0.9841666666666666,
|
| 577 |
-
"acc_at_50": 0.
|
| 578 |
"chance": 0.5,
|
| 579 |
-
"mean_conf": 0.
|
| 580 |
-
"sec":
|
| 581 |
"heldout": false
|
| 582 |
},
|
| 583 |
"qqp": {
|
| 584 |
"n": 1500,
|
| 585 |
-
"acc": 0.
|
| 586 |
-
"nll": 0.
|
| 587 |
-
"brier": 0.
|
| 588 |
-
"ece": 0.
|
| 589 |
-
"aurc": 0.
|
| 590 |
-
"acc_at_80": 0.
|
| 591 |
-
"acc_at_50": 0.
|
| 592 |
"chance": 0.5,
|
| 593 |
-
"mean_conf": 0.
|
| 594 |
-
"sec": 2.
|
| 595 |
"heldout": false
|
| 596 |
},
|
| 597 |
"mrpc": {
|
| 598 |
"n": 408,
|
| 599 |
"acc": 0.8455882352941176,
|
| 600 |
-
"nll": 0.
|
| 601 |
-
"brier": 0.
|
| 602 |
-
"ece": 0.
|
| 603 |
-
"aurc": 0.
|
| 604 |
-
"acc_at_80": 0.
|
| 605 |
"acc_at_50": 0.9803921568627451,
|
| 606 |
"chance": 0.5,
|
| 607 |
-
"mean_conf": 0.
|
| 608 |
-
"sec": 0.
|
| 609 |
"heldout": false
|
| 610 |
},
|
| 611 |
"paws": {
|
| 612 |
"n": 1500,
|
| 613 |
-
"acc": 0.
|
| 614 |
-
"nll": 0.
|
| 615 |
-
"brier": 0.
|
| 616 |
-
"ece": 0.
|
| 617 |
-
"aurc": 0.
|
| 618 |
-
"acc_at_80": 0.
|
| 619 |
-
"acc_at_50": 0.
|
| 620 |
"chance": 0.5,
|
| 621 |
-
"mean_conf": 0.
|
| 622 |
-
"sec":
|
| 623 |
"heldout": true
|
| 624 |
},
|
| 625 |
"cola": {
|
| 626 |
"n": 1043,
|
| 627 |
-
"acc": 0.
|
| 628 |
-
"nll": 0.
|
| 629 |
-
"brier": 0.
|
| 630 |
-
"ece": 0.
|
| 631 |
-
"aurc": 0.
|
| 632 |
-
"acc_at_80": 0.
|
| 633 |
"acc_at_50": 0.9367816091954023,
|
| 634 |
"chance": 0.5,
|
| 635 |
-
"mean_conf": 0.
|
| 636 |
-
"sec": 1.
|
| 637 |
"heldout": false
|
| 638 |
},
|
| 639 |
"boolq": {
|
| 640 |
"n": 1500,
|
| 641 |
-
"acc": 0.
|
| 642 |
-
"nll": 0.
|
| 643 |
-
"brier": 0.
|
| 644 |
-
"ece": 0.
|
| 645 |
-
"aurc": 0.
|
| 646 |
-
"acc_at_80": 0.
|
| 647 |
-
"acc_at_50": 0.
|
| 648 |
"chance": 0.5,
|
| 649 |
-
"mean_conf": 0.
|
| 650 |
-
"sec":
|
| 651 |
"heldout": false
|
| 652 |
},
|
| 653 |
"strategyqa": {
|
| 654 |
"n": 687,
|
| 655 |
-
"acc": 0.
|
| 656 |
-
"nll": 0.
|
| 657 |
-
"brier": 0.
|
| 658 |
-
"ece": 0.
|
| 659 |
-
"aurc": 0.
|
| 660 |
-
"acc_at_80": 0.
|
| 661 |
-
"acc_at_50": 0.
|
| 662 |
"chance": 0.5,
|
| 663 |
-
"mean_conf": 0.
|
| 664 |
-
"sec":
|
| 665 |
"heldout": true
|
| 666 |
},
|
| 667 |
"pubmedqa": {
|
| 668 |
"n": 500,
|
| 669 |
-
"acc": 0.
|
| 670 |
-
"nll": 0.
|
| 671 |
-
"brier": 0.
|
| 672 |
-
"ece": 0.
|
| 673 |
-
"aurc": 0.
|
| 674 |
-
"acc_at_80": 0.
|
| 675 |
-
"acc_at_50": 0.
|
| 676 |
"chance": 0.33333333333333326,
|
| 677 |
-
"mean_conf": 0.
|
| 678 |
-
"sec": 2.
|
| 679 |
"heldout": true
|
| 680 |
},
|
| 681 |
"arc": {
|
| 682 |
"n": 1172,
|
| 683 |
-
"acc": 0.
|
| 684 |
-
"nll": 0.
|
| 685 |
-
"brier": 0.
|
| 686 |
-
"ece": 0.
|
| 687 |
-
"aurc": 0.
|
| 688 |
-
"acc_at_80": 0.
|
| 689 |
-
"acc_at_50": 0.
|
| 690 |
"chance": 0.25015642775881686,
|
| 691 |
-
"mean_conf": 0.
|
| 692 |
-
"sec":
|
| 693 |
"heldout": false
|
| 694 |
},
|
| 695 |
"commonsense_qa": {
|
| 696 |
"n": 1221,
|
| 697 |
-
"acc": 0.
|
| 698 |
-
"nll": 0.
|
| 699 |
-
"brier": 0.
|
| 700 |
-
"ece": 0.
|
| 701 |
-
"aurc": 0.
|
| 702 |
-
"acc_at_80": 0.
|
| 703 |
-
"acc_at_50": 0.
|
| 704 |
"chance": 0.19999999999999998,
|
| 705 |
-
"mean_conf": 0.
|
| 706 |
"sec": 2.0,
|
| 707 |
"heldout": false
|
| 708 |
},
|
| 709 |
"qasc": {
|
| 710 |
"n": 926,
|
| 711 |
-
"acc": 0.
|
| 712 |
-
"nll": 0.
|
| 713 |
-
"brier": 0.
|
| 714 |
-
"ece": 0.
|
| 715 |
-
"aurc": 0.
|
| 716 |
-
"acc_at_80": 0.
|
| 717 |
-
"acc_at_50": 0.
|
| 718 |
"chance": 0.125,
|
| 719 |
-
"mean_conf": 0.
|
| 720 |
-
"sec": 1.
|
| 721 |
"heldout": false
|
| 722 |
},
|
| 723 |
"openbookqa": {
|
| 724 |
"n": 500,
|
| 725 |
-
"acc": 0.
|
| 726 |
-
"nll": 0.
|
| 727 |
-
"brier": 0.
|
| 728 |
-
"ece": 0.
|
| 729 |
-
"aurc": 0.
|
| 730 |
"acc_at_80": 0.905,
|
| 731 |
-
"acc_at_50": 0.
|
| 732 |
"chance": 0.25,
|
| 733 |
-
"mean_conf": 0.
|
| 734 |
-
"sec": 0.
|
| 735 |
"heldout": false
|
| 736 |
},
|
| 737 |
"sciq": {
|
| 738 |
"n": 1000,
|
| 739 |
-
"acc": 0.
|
| 740 |
-
"nll": 0.
|
| 741 |
-
"brier": 0.
|
| 742 |
-
"ece": 0.
|
| 743 |
-
"aurc": 0.
|
| 744 |
"acc_at_80": 1.0,
|
| 745 |
"acc_at_50": 1.0,
|
| 746 |
"chance": 0.25,
|
| 747 |
-
"mean_conf": 0.
|
| 748 |
-
"sec":
|
| 749 |
"heldout": true
|
| 750 |
},
|
| 751 |
"hellaswag": {
|
| 752 |
"n": 1500,
|
| 753 |
-
"acc": 0.
|
| 754 |
-
"nll": 0.
|
| 755 |
-
"brier": 0.
|
| 756 |
-
"ece": 0.
|
| 757 |
-
"aurc": 0.
|
| 758 |
-
"acc_at_80": 0.
|
| 759 |
"acc_at_50": 0.9986666666666667,
|
| 760 |
"chance": 0.25,
|
| 761 |
-
"mean_conf": 0.
|
| 762 |
-
"sec": 4.
|
| 763 |
"heldout": false
|
| 764 |
},
|
| 765 |
"piqa": {
|
| 766 |
"n": 1500,
|
| 767 |
-
"acc": 0.
|
| 768 |
-
"nll": 0.
|
| 769 |
-
"brier": 0.
|
| 770 |
-
"ece": 0.
|
| 771 |
-
"aurc": 0.
|
| 772 |
-
"acc_at_80": 0.
|
| 773 |
-
"acc_at_50": 0.
|
| 774 |
"chance": 0.5,
|
| 775 |
-
"mean_conf": 0.
|
| 776 |
-
"sec": 2.
|
| 777 |
"heldout": false
|
| 778 |
},
|
| 779 |
"social_iqa": {
|
| 780 |
"n": 1500,
|
| 781 |
-
"acc": 0.
|
| 782 |
-
"nll": 0.
|
| 783 |
-
"brier": 0.
|
| 784 |
-
"ece": 0.
|
| 785 |
-
"aurc": 0.
|
| 786 |
-
"acc_at_80": 0.
|
| 787 |
-
"acc_at_50": 0.
|
| 788 |
"chance": 0.3333333333333333,
|
| 789 |
-
"mean_conf": 0.
|
| 790 |
-
"sec": 2.
|
| 791 |
"heldout": true
|
| 792 |
},
|
| 793 |
"winogrande": {
|
| 794 |
"n": 1267,
|
| 795 |
-
"acc": 0.
|
| 796 |
-
"nll": 0.
|
| 797 |
-
"brier": 0.
|
| 798 |
-
"ece": 0.
|
| 799 |
-
"aurc": 0.
|
| 800 |
-
"acc_at_80": 0.
|
| 801 |
-
"acc_at_50": 0.
|
| 802 |
"chance": 0.5,
|
| 803 |
-
"mean_conf": 0.
|
| 804 |
-
"sec":
|
| 805 |
"heldout": false
|
| 806 |
},
|
| 807 |
"race": {
|
| 808 |
"n": 1500,
|
| 809 |
-
"acc": 0.
|
| 810 |
-
"nll": 0.
|
| 811 |
-
"brier": 0.
|
| 812 |
-
"ece": 0.
|
| 813 |
-
"aurc": 0.
|
| 814 |
-
"acc_at_80": 0.
|
| 815 |
-
"acc_at_50": 0.
|
| 816 |
"chance": 0.25,
|
| 817 |
-
"mean_conf": 0.
|
| 818 |
-
"sec": 7.
|
| 819 |
"heldout": false
|
| 820 |
},
|
| 821 |
"mmlu": {
|
| 822 |
"n": 1500,
|
| 823 |
-
"acc": 0.
|
| 824 |
-
"nll": 0.
|
| 825 |
-
"brier": 0.
|
| 826 |
-
"ece": 0.
|
| 827 |
-
"aurc": 0.
|
| 828 |
-
"acc_at_80": 0.
|
| 829 |
-
"acc_at_50": 0.
|
| 830 |
"chance": 0.25,
|
| 831 |
-
"mean_conf": 0.
|
| 832 |
-
"sec": 3.
|
| 833 |
"heldout": false
|
| 834 |
},
|
| 835 |
"medqa": {
|
| 836 |
"n": 1273,
|
| 837 |
-
"acc": 0.
|
| 838 |
-
"nll": 1.
|
| 839 |
-
"brier": 0.
|
| 840 |
-
"ece": 0.
|
| 841 |
-
"aurc": 0.
|
| 842 |
-
"acc_at_80": 0.
|
| 843 |
-
"acc_at_50": 0.
|
| 844 |
"chance": 0.25,
|
| 845 |
-
"mean_conf": 0.
|
| 846 |
-
"sec": 4.
|
| 847 |
"heldout": false
|
| 848 |
},
|
| 849 |
"truthfulqa": {
|
| 850 |
"n": 817,
|
| 851 |
-
"acc": 0.
|
| 852 |
-
"nll": 1.
|
| 853 |
-
"brier": 0.
|
| 854 |
-
"ece": 0.
|
| 855 |
-
"aurc": 0.
|
| 856 |
-
"acc_at_80": 0.
|
| 857 |
-
"acc_at_50": 0.
|
| 858 |
"chance": 0.22622350449767833,
|
| 859 |
-
"mean_conf": 0.
|
| 860 |
-
"sec": 1.
|
| 861 |
"heldout": true
|
| 862 |
},
|
| 863 |
"fin_phrasebank": {
|
| 864 |
"n": 970,
|
| 865 |
-
"acc": 0.
|
| 866 |
-
"nll": 0.
|
| 867 |
-
"brier": 0.
|
| 868 |
-
"ece": 0.
|
| 869 |
-
"aurc": 0.
|
| 870 |
-
"acc_at_80": 0.
|
| 871 |
-
"acc_at_50": 0.
|
| 872 |
"chance": 0.33333333333333326,
|
| 873 |
-
"mean_conf": 0.
|
| 874 |
-
"sec":
|
| 875 |
"heldout": true
|
| 876 |
},
|
| 877 |
"banking77": {
|
| 878 |
"n": 1500,
|
| 879 |
-
"acc": 0.
|
| 880 |
-
"nll": 0.
|
| 881 |
-
"brier": 0.
|
| 882 |
-
"ece": 0.
|
| 883 |
-
"aurc": 0.
|
| 884 |
"acc_at_80": 0.9966666666666667,
|
| 885 |
"acc_at_50": 1.0,
|
| 886 |
"chance": 0.1,
|
| 887 |
-
"mean_conf": 0.
|
| 888 |
-
"sec":
|
| 889 |
"heldout": false
|
| 890 |
},
|
| 891 |
"bias_in_bios": {
|
| 892 |
"n": 3000,
|
| 893 |
-
"acc": 0.
|
| 894 |
-
"nll": 0.
|
| 895 |
-
"brier": 0.
|
| 896 |
-
"ece": 0.
|
| 897 |
-
"aurc": 0.
|
| 898 |
-
"acc_at_80": 0.
|
| 899 |
"acc_at_50": 1.0,
|
| 900 |
"chance": 0.30000000000000004,
|
| 901 |
-
"mean_conf": 0.
|
| 902 |
-
"sec":
|
| 903 |
"heldout": false,
|
| 904 |
"per_q": [
|
| 905 |
-
0.
|
| 906 |
0.9973333333333333
|
| 907 |
]
|
| 908 |
},
|
| 909 |
"helpsteer2": {
|
| 910 |
"n": 5190,
|
| 911 |
-
"acc": 0.
|
| 912 |
-
"nll": 0.
|
| 913 |
-
"brier": 0.
|
| 914 |
-
"ece": 0.
|
| 915 |
-
"aurc": 0.
|
| 916 |
-
"acc_at_80": 0.
|
| 917 |
-
"acc_at_50": 0.
|
| 918 |
"chance": 0.2,
|
| 919 |
-
"mean_conf": 0.
|
| 920 |
-
"sec":
|
| 921 |
"heldout": false,
|
| 922 |
"per_q": [
|
| 923 |
-
0.
|
| 924 |
-
0.
|
| 925 |
-
0.
|
| 926 |
-
0.
|
| 927 |
-
0.
|
| 928 |
]
|
| 929 |
},
|
| 930 |
"helpsteer3_pref": {
|
| 931 |
"n": 1176,
|
| 932 |
-
"acc": 0.
|
| 933 |
-
"nll": 1.
|
| 934 |
-
"brier": 0.
|
| 935 |
-
"ece": 0.
|
| 936 |
-
"aurc": 0.
|
| 937 |
-
"acc_at_80": 0.
|
| 938 |
-
"acc_at_50": 0.
|
| 939 |
"chance": 0.14285714285714282,
|
| 940 |
-
"mean_conf": 0.
|
| 941 |
-
"sec":
|
| 942 |
"heldout": false
|
| 943 |
},
|
| 944 |
"hate_speech_scales": {
|
| 945 |
"n": 7500,
|
| 946 |
-
"acc": 0.
|
| 947 |
-
"nll": 0.
|
| 948 |
-
"brier": 0.
|
| 949 |
-
"ece": 0.
|
| 950 |
-
"aurc": 0.
|
| 951 |
-
"acc_at_80": 0.
|
| 952 |
-
"acc_at_50": 0.
|
| 953 |
"chance": 0.22666666666666674,
|
| 954 |
-
"mean_conf": 0.
|
| 955 |
-
"sec":
|
| 956 |
"heldout": false,
|
| 957 |
"per_q": [
|
| 958 |
0.706,
|
| 959 |
-
0.
|
| 960 |
-
0.
|
| 961 |
-
0.
|
| 962 |
-
0.
|
| 963 |
]
|
| 964 |
},
|
| 965 |
"liar2": {
|
| 966 |
"n": 1500,
|
| 967 |
-
"acc": 0.
|
| 968 |
-
"nll": 1.
|
| 969 |
-
"brier": 0.
|
| 970 |
-
"ece": 0.
|
| 971 |
-
"aurc": 0.
|
| 972 |
-
"acc_at_80": 0.
|
| 973 |
-
"acc_at_50": 0.
|
| 974 |
"chance": 0.16666666666666666,
|
| 975 |
-
"mean_conf": 0.
|
| 976 |
-
"sec": 3.
|
| 977 |
"heldout": false
|
| 978 |
},
|
| 979 |
"prosocial_safety": {
|
| 980 |
"n": 1500,
|
| 981 |
-
"acc": 0.
|
| 982 |
-
"nll": 1.
|
| 983 |
-
"brier": 0.
|
| 984 |
-
"ece": 0.
|
| 985 |
-
"aurc": 0.
|
| 986 |
-
"acc_at_80": 0.
|
| 987 |
-
"acc_at_50": 0.
|
| 988 |
"chance": 0.2,
|
| 989 |
-
"mean_conf": 0.
|
| 990 |
-
"sec":
|
| 991 |
"heldout": false
|
| 992 |
},
|
| 993 |
"ultrafeedback_pref": {
|
| 994 |
"n": 1500,
|
| 995 |
-
"acc": 0.
|
| 996 |
-
"nll": 0.
|
| 997 |
-
"brier": 0.
|
| 998 |
-
"ece": 0.
|
| 999 |
-
"aurc": 0.
|
| 1000 |
-
"acc_at_80": 0.
|
| 1001 |
-
"acc_at_50": 0.
|
| 1002 |
"chance": 0.5,
|
| 1003 |
-
"mean_conf": 0.
|
| 1004 |
-
"sec":
|
| 1005 |
"heldout": false
|
| 1006 |
},
|
| 1007 |
"shp": {
|
| 1008 |
"n": 1500,
|
| 1009 |
-
"acc": 0.
|
| 1010 |
-
"nll": 0.
|
| 1011 |
-
"brier": 0.
|
| 1012 |
-
"ece": 0.
|
| 1013 |
-
"aurc": 0.
|
| 1014 |
-
"acc_at_80": 0.
|
| 1015 |
-
"acc_at_50": 0.
|
| 1016 |
"chance": 0.5,
|
| 1017 |
-
"mean_conf": 0.
|
| 1018 |
-
"sec": 7.
|
| 1019 |
"heldout": false
|
| 1020 |
},
|
| 1021 |
"hh_rlhf": {
|
| 1022 |
"n": 1488,
|
| 1023 |
-
"acc": 0.
|
| 1024 |
-
"nll": 0.
|
| 1025 |
-
"brier": 0.
|
| 1026 |
-
"ece": 0.
|
| 1027 |
-
"aurc": 0.
|
| 1028 |
-
"acc_at_80": 0.
|
| 1029 |
-
"acc_at_50": 0.
|
| 1030 |
"chance": 0.5,
|
| 1031 |
-
"mean_conf": 0.
|
| 1032 |
-
"sec":
|
| 1033 |
"heldout": false
|
| 1034 |
},
|
| 1035 |
"arena_pref": {
|
| 1036 |
"n": 1500,
|
| 1037 |
-
"acc": 0.
|
| 1038 |
-
"nll": 1.
|
| 1039 |
-
"brier": 0.
|
| 1040 |
-
"ece": 0.
|
| 1041 |
-
"aurc": 0.
|
| 1042 |
-
"acc_at_80": 0.
|
| 1043 |
-
"acc_at_50": 0.
|
| 1044 |
"chance": 0.3333333333333333,
|
| 1045 |
-
"mean_conf": 0.
|
| 1046 |
-
"sec":
|
| 1047 |
"heldout": true
|
| 1048 |
},
|
| 1049 |
"reward_bench": {
|
| 1050 |
"n": 1500,
|
| 1051 |
-
"acc": 0.
|
| 1052 |
-
"nll": 0.
|
| 1053 |
-
"brier": 0.
|
| 1054 |
-
"ece": 0.
|
| 1055 |
-
"aurc": 0.
|
| 1056 |
-
"acc_at_80": 0.
|
| 1057 |
-
"acc_at_50": 0.
|
| 1058 |
"chance": 0.5,
|
| 1059 |
-
"mean_conf": 0.
|
| 1060 |
-
"sec":
|
| 1061 |
"heldout": true
|
| 1062 |
},
|
| 1063 |
"glaive_tools": {
|
| 1064 |
"n": 1348,
|
| 1065 |
-
"acc": 0.
|
| 1066 |
-
"nll": 0.
|
| 1067 |
-
"brier": 0.
|
| 1068 |
-
"ece": 0.
|
| 1069 |
-
"aurc": 0.
|
| 1070 |
-
"acc_at_80": 0.
|
| 1071 |
-
"acc_at_50": 0.
|
| 1072 |
"chance": 0.16666666666666666,
|
| 1073 |
-
"mean_conf": 0.
|
| 1074 |
-
"sec":
|
| 1075 |
"heldout": false
|
| 1076 |
},
|
| 1077 |
"hermes_tools": {
|
| 1078 |
"n": 1500,
|
| 1079 |
-
"acc": 0.
|
| 1080 |
-
"nll": 0.
|
| 1081 |
-
"brier": 0.
|
| 1082 |
-
"ece": 0.
|
| 1083 |
-
"aurc": 0.
|
| 1084 |
-
"acc_at_80": 0.
|
| 1085 |
-
"acc_at_50": 0.
|
| 1086 |
"chance": 0.125,
|
| 1087 |
-
"mean_conf": 0.
|
| 1088 |
-
"sec":
|
| 1089 |
"heldout": true
|
| 1090 |
},
|
| 1091 |
"copa": {
|
| 1092 |
"n": 100,
|
| 1093 |
-
"acc": 0.
|
| 1094 |
-
"nll": 0.
|
| 1095 |
-
"brier": 0.
|
| 1096 |
-
"ece": 0.
|
| 1097 |
-
"aurc": 0.
|
| 1098 |
"acc_at_80": 0.975,
|
| 1099 |
"acc_at_50": 0.98,
|
| 1100 |
"chance": 0.5,
|
| 1101 |
-
"mean_conf": 0.
|
| 1102 |
-
"sec":
|
| 1103 |
"heldout": false
|
| 1104 |
},
|
| 1105 |
"wic": {
|
| 1106 |
"n": 638,
|
| 1107 |
-
"acc": 0.
|
| 1108 |
-
"nll": 0.
|
| 1109 |
-
"brier": 0.
|
| 1110 |
-
"ece": 0.
|
| 1111 |
-
"aurc": 0.
|
| 1112 |
-
"acc_at_80": 0.
|
| 1113 |
-
"acc_at_50": 0.
|
| 1114 |
"chance": 0.5,
|
| 1115 |
-
"mean_conf": 0.
|
| 1116 |
-
"sec": 1.
|
| 1117 |
"heldout": false
|
| 1118 |
},
|
| 1119 |
"multirc": {
|
| 1120 |
"n": 1500,
|
| 1121 |
-
"acc": 0.
|
| 1122 |
-
"nll": 0.
|
| 1123 |
-
"brier": 0.
|
| 1124 |
-
"ece": 0.
|
| 1125 |
-
"aurc": 0.
|
| 1126 |
-
"acc_at_80": 0.
|
| 1127 |
-
"acc_at_50": 0.
|
| 1128 |
"chance": 0.5,
|
| 1129 |
-
"mean_conf": 0.
|
| 1130 |
-
"sec":
|
| 1131 |
"heldout": false
|
| 1132 |
},
|
| 1133 |
"cb": {
|
| 1134 |
"n": 56,
|
| 1135 |
"acc": 0.8928571428571429,
|
| 1136 |
-
"nll": 0.
|
| 1137 |
-
"brier": 0.
|
| 1138 |
-
"ece": 0.
|
| 1139 |
-
"aurc": 0.
|
| 1140 |
-
"acc_at_80": 0.
|
| 1141 |
"acc_at_50": 0.9642857142857143,
|
| 1142 |
"chance": 0.3333333333333333,
|
| 1143 |
-
"mean_conf": 0.
|
| 1144 |
"sec": 0.2,
|
| 1145 |
"heldout": true
|
| 1146 |
},
|
| 1147 |
"fever": {
|
| 1148 |
"n": 1500,
|
| 1149 |
-
"acc": 0.
|
| 1150 |
-
"nll": 0.
|
| 1151 |
-
"brier": 0.
|
| 1152 |
-
"ece": 0.
|
| 1153 |
-
"aurc": 0.
|
| 1154 |
-
"acc_at_80": 0.
|
| 1155 |
-
"acc_at_50": 0.
|
| 1156 |
"chance": 0.3333333333333333,
|
| 1157 |
-
"mean_conf": 0.
|
| 1158 |
-
"sec":
|
| 1159 |
"heldout": false
|
| 1160 |
},
|
| 1161 |
"wiki_qa": {
|
| 1162 |
"n": 879,
|
| 1163 |
-
"acc": 0.
|
| 1164 |
-
"nll": 0.
|
| 1165 |
-
"brier": 0.
|
| 1166 |
-
"ece": 0.
|
| 1167 |
-
"aurc": 0.
|
| 1168 |
-
"acc_at_80": 0.
|
| 1169 |
-
"acc_at_50": 0.
|
| 1170 |
"chance": 0.5,
|
| 1171 |
-
"mean_conf": 0.
|
| 1172 |
-
"sec":
|
| 1173 |
"heldout": false
|
| 1174 |
},
|
| 1175 |
"msmarco_rel": {
|
| 1176 |
"n": 1446,
|
| 1177 |
-
"acc": 0.
|
| 1178 |
-
"nll": 0.
|
| 1179 |
-
"brier": 0.
|
| 1180 |
-
"ece": 0.
|
| 1181 |
-
"aurc": 0.
|
| 1182 |
-
"acc_at_80": 0.
|
| 1183 |
"acc_at_50": 0.7745504840940526,
|
| 1184 |
"chance": 0.5,
|
| 1185 |
-
"mean_conf": 0.
|
| 1186 |
-
"sec": 3.
|
| 1187 |
"heldout": false
|
| 1188 |
},
|
| 1189 |
"medmcqa": {
|
| 1190 |
"n": 1500,
|
| 1191 |
-
"acc": 0.
|
| 1192 |
-
"nll": 1.
|
| 1193 |
-
"brier": 0.
|
| 1194 |
-
"ece": 0.
|
| 1195 |
-
"aurc": 0.
|
| 1196 |
-
"acc_at_80": 0.
|
| 1197 |
-
"acc_at_50": 0.
|
| 1198 |
"chance": 0.25,
|
| 1199 |
-
"mean_conf": 0.
|
| 1200 |
-
"sec": 2.
|
| 1201 |
"heldout": false
|
| 1202 |
},
|
| 1203 |
"quality": {
|
| 1204 |
"n": 1500,
|
| 1205 |
-
"acc": 0.
|
| 1206 |
-
"nll": 1.
|
| 1207 |
-
"brier": 0.
|
| 1208 |
-
"ece": 0.
|
| 1209 |
-
"aurc": 0.
|
| 1210 |
-
"acc_at_80": 0.
|
| 1211 |
-
"acc_at_50": 0.
|
| 1212 |
"chance": 0.25,
|
| 1213 |
-
"mean_conf": 0.
|
| 1214 |
-
"sec":
|
| 1215 |
"heldout": true
|
| 1216 |
},
|
| 1217 |
"xstory_cloze": {
|
| 1218 |
"n": 1500,
|
| 1219 |
-
"acc": 0.
|
| 1220 |
-
"nll": 0.
|
| 1221 |
-
"brier": 0.
|
| 1222 |
-
"ece": 0.
|
| 1223 |
-
"aurc": 0.
|
| 1224 |
-
"acc_at_80": 0.
|
| 1225 |
"acc_at_50": 0.9986666666666667,
|
| 1226 |
"chance": 0.5,
|
| 1227 |
-
"mean_conf": 0.
|
| 1228 |
-
"sec":
|
| 1229 |
"heldout": true
|
| 1230 |
},
|
| 1231 |
"abstain_probe": {
|
| 1232 |
"n": 1500,
|
| 1233 |
-
"acc": 0.
|
| 1234 |
-
"nll": 0.
|
| 1235 |
-
"brier": 0.
|
| 1236 |
-
"ece": 0.
|
| 1237 |
-
"aurc": 0.
|
| 1238 |
-
"acc_at_80": 0.
|
| 1239 |
-
"acc_at_50": 0.
|
| 1240 |
"chance": 0.17950317460317464,
|
| 1241 |
-
"mean_conf": 0.
|
| 1242 |
-
"sec": 3.
|
| 1243 |
"heldout": true
|
| 1244 |
},
|
| 1245 |
"toolace": {
|
| 1246 |
"n": 1000,
|
| 1247 |
-
"acc": 0.
|
| 1248 |
-
"nll": 0.
|
| 1249 |
-
"brier": 0.
|
| 1250 |
-
"ece": 0.
|
| 1251 |
-
"aurc": 0.
|
| 1252 |
"acc_at_80": 0.97125,
|
| 1253 |
-
"acc_at_50": 0.
|
| 1254 |
"chance": 0.12470555555555554,
|
| 1255 |
-
"mean_conf": 0.
|
| 1256 |
-
"sec":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1257 |
"heldout": false
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1258 |
}
|
| 1259 |
},
|
| 1260 |
"agg": {
|
| 1261 |
"in_task": {
|
| 1262 |
-
"acc": 0.
|
| 1263 |
-
"nll": 0.
|
| 1264 |
-
"brier": 0.
|
| 1265 |
-
"ece": 0.
|
| 1266 |
-
"aurc": 0.
|
| 1267 |
-
"acc_at_80": 0.
|
| 1268 |
-
"chance": 0.
|
| 1269 |
},
|
| 1270 |
"heldout": {
|
| 1271 |
-
"acc": 0.
|
| 1272 |
-
"nll": 0.
|
| 1273 |
-
"brier": 0.
|
| 1274 |
-
"ece": 0.
|
| 1275 |
-
"aurc": 0.
|
| 1276 |
-
"acc_at_80": 0.
|
| 1277 |
"chance": 0.32053884112032693
|
| 1278 |
},
|
| 1279 |
-
"n_in":
|
| 1280 |
"n_heldout": 23
|
| 1281 |
},
|
| 1282 |
-
"model": "runs/
|
|
|
|
| 1283 |
}
|
|
|
|
| 2 |
"results": {
|
| 3 |
"clinc_oos": {
|
| 4 |
"n": 1500,
|
| 5 |
+
"acc": 0.978,
|
| 6 |
+
"nll": 0.06232544407248497,
|
| 7 |
+
"brier": 0.033815909177064896,
|
| 8 |
+
"ece": 0.008496353924274395,
|
| 9 |
+
"aurc": 0.0008730387999259201,
|
| 10 |
"acc_at_80": 0.9991666666666666,
|
| 11 |
"acc_at_50": 1.0,
|
| 12 |
"chance": 0.1,
|
| 13 |
+
"mean_conf": 0.9839518070220947,
|
| 14 |
+
"sec": 13.8,
|
| 15 |
"heldout": false
|
| 16 |
},
|
| 17 |
"massive_intent": {
|
| 18 |
"n": 1500,
|
| 19 |
+
"acc": 0.9486666666666667,
|
| 20 |
+
"nll": 0.15551087260246277,
|
| 21 |
+
"brier": 0.0762505829334259,
|
| 22 |
+
"ece": 0.012290600876013396,
|
| 23 |
+
"aurc": 0.005614506940959013,
|
| 24 |
"acc_at_80": 0.9933333333333333,
|
| 25 |
+
"acc_at_50": 0.9973333333333333,
|
| 26 |
"chance": 0.1,
|
| 27 |
+
"mean_conf": 0.953382670879364,
|
| 28 |
+
"sec": 4.0,
|
| 29 |
"heldout": false
|
| 30 |
},
|
| 31 |
"massive_scenario": {
|
| 32 |
"n": 1500,
|
| 33 |
+
"acc": 0.8033333333333333,
|
| 34 |
+
"nll": 0.5681269764900208,
|
| 35 |
+
"brier": 0.2758673429489136,
|
| 36 |
+
"ece": 0.03523811907569568,
|
| 37 |
+
"aurc": 0.05579813486552701,
|
| 38 |
+
"acc_at_80": 0.8858333333333334,
|
| 39 |
+
"acc_at_50": 0.968,
|
| 40 |
"chance": 0.1,
|
| 41 |
+
"mean_conf": 0.8278322219848633,
|
| 42 |
+
"sec": 1.5,
|
| 43 |
"heldout": true
|
| 44 |
},
|
| 45 |
"bitext_support": {
|
| 46 |
"n": 3000,
|
| 47 |
+
"acc": 0.998,
|
| 48 |
+
"nll": 0.007169861346483231,
|
| 49 |
+
"brier": 0.0035593616776168346,
|
| 50 |
+
"ece": 0.0009388573169708132,
|
| 51 |
+
"aurc": 1.788459220568468e-05,
|
| 52 |
"acc_at_80": 1.0,
|
| 53 |
"acc_at_50": 1.0,
|
| 54 |
"chance": 0.09999999999999998,
|
| 55 |
+
"mean_conf": 0.998285710811615,
|
| 56 |
+
"sec": 2.1,
|
| 57 |
"heldout": false,
|
| 58 |
"per_q": [
|
| 59 |
+
0.9986666666666667,
|
| 60 |
+
0.9973333333333333
|
| 61 |
]
|
| 62 |
},
|
| 63 |
"support_tickets": {
|
| 64 |
"n": 4500,
|
| 65 |
+
"acc": 0.5737777777777778,
|
| 66 |
+
"nll": 0.9863417148590088,
|
| 67 |
+
"brier": 0.5184141993522644,
|
| 68 |
+
"ece": 0.0138452862434917,
|
| 69 |
+
"aurc": 0.21858182796967993,
|
| 70 |
+
"acc_at_80": 0.6361111111111111,
|
| 71 |
+
"acc_at_50": 0.7497777777777778,
|
| 72 |
"chance": 0.22777777777777777,
|
| 73 |
+
"mean_conf": 0.5744046568870544,
|
| 74 |
+
"sec": 10.3,
|
| 75 |
"heldout": false,
|
| 76 |
"per_q": [
|
| 77 |
+
0.8113333333333334,
|
| 78 |
+
0.43133333333333335,
|
| 79 |
+
0.4786666666666667
|
| 80 |
]
|
| 81 |
},
|
| 82 |
"ag_news": {
|
| 83 |
"n": 1500,
|
| 84 |
"acc": 0.9173333333333333,
|
| 85 |
+
"nll": 0.24114005267620087,
|
| 86 |
+
"brier": 0.12489213049411774,
|
| 87 |
+
"ece": 0.016586370726426467,
|
| 88 |
+
"aurc": 0.018462071969618875,
|
| 89 |
"acc_at_80": 0.9716666666666667,
|
| 90 |
+
"acc_at_50": 0.9853333333333333,
|
| 91 |
"chance": 0.25,
|
| 92 |
+
"mean_conf": 0.9118769764900208,
|
| 93 |
+
"sec": 1.5,
|
| 94 |
"heldout": false
|
| 95 |
},
|
| 96 |
"dbpedia": {
|
| 97 |
"n": 1500,
|
| 98 |
"acc": 0.9913333333333333,
|
| 99 |
+
"nll": 0.024175753816962242,
|
| 100 |
+
"brier": 0.013113866560161114,
|
| 101 |
+
"ece": 0.005916546404361727,
|
| 102 |
+
"aurc": 0.00010674119407793753,
|
| 103 |
"acc_at_80": 1.0,
|
| 104 |
"acc_at_50": 1.0,
|
| 105 |
"chance": 0.1,
|
| 106 |
+
"mean_conf": 0.9923719763755798,
|
| 107 |
+
"sec": 2.0,
|
| 108 |
"heldout": false
|
| 109 |
},
|
| 110 |
"yahoo_topics": {
|
| 111 |
"n": 1500,
|
| 112 |
+
"acc": 0.76,
|
| 113 |
+
"nll": 0.7366493940353394,
|
| 114 |
+
"brier": 0.33801746368408203,
|
| 115 |
+
"ece": 0.028587430814902,
|
| 116 |
+
"aurc": 0.08282859840202575,
|
| 117 |
+
"acc_at_80": 0.855,
|
| 118 |
+
"acc_at_50": 0.9453333333333334,
|
| 119 |
"chance": 0.1,
|
| 120 |
+
"mean_conf": 0.7742340564727783,
|
| 121 |
+
"sec": 1.8,
|
| 122 |
"heldout": false
|
| 123 |
},
|
| 124 |
"newsgroups": {
|
| 125 |
"n": 1462,
|
| 126 |
+
"acc": 0.8125854993160054,
|
| 127 |
+
"nll": 0.5523399710655212,
|
| 128 |
+
"brier": 0.26114463806152344,
|
| 129 |
+
"ece": 0.044583431815807525,
|
| 130 |
+
"aurc": 0.04188347828916838,
|
| 131 |
"acc_at_80": 0.917094017094017,
|
| 132 |
"acc_at_50": 0.9835841313269493,
|
| 133 |
"chance": 0.09999999999999999,
|
| 134 |
+
"mean_conf": 0.8485810160636902,
|
| 135 |
+
"sec": 14.4,
|
| 136 |
"heldout": false
|
| 137 |
},
|
| 138 |
"bbc_news": {
|
| 139 |
"n": 1000,
|
| 140 |
+
"acc": 0.939,
|
| 141 |
+
"nll": 0.16493572294712067,
|
| 142 |
+
"brier": 0.09448844194412231,
|
| 143 |
+
"ece": 0.01869928520917896,
|
| 144 |
+
"aurc": 0.008020306999373308,
|
| 145 |
+
"acc_at_80": 0.9825,
|
| 146 |
"acc_at_50": 1.0,
|
| 147 |
"chance": 0.20000000000000004,
|
| 148 |
+
"mean_conf": 0.9388766884803772,
|
| 149 |
+
"sec": 5.8,
|
| 150 |
"heldout": true
|
| 151 |
},
|
| 152 |
"trec": {
|
| 153 |
"n": 500,
|
| 154 |
+
"acc": 0.754,
|
| 155 |
+
"nll": 0.6227552890777588,
|
| 156 |
+
"brier": 0.33451518416404724,
|
| 157 |
+
"ece": 0.041722716212272644,
|
| 158 |
+
"aurc": 0.10665603039652902,
|
| 159 |
+
"acc_at_80": 0.82,
|
| 160 |
+
"acc_at_50": 0.896,
|
| 161 |
"chance": 0.16666666666666663,
|
| 162 |
+
"mean_conf": 0.7837464213371277,
|
| 163 |
+
"sec": 3.3,
|
| 164 |
"heldout": true
|
| 165 |
},
|
| 166 |
"student_questions": {
|
| 167 |
"n": 1500,
|
| 168 |
+
"acc": 0.9246666666666666,
|
| 169 |
+
"nll": 0.22754858434200287,
|
| 170 |
+
"brier": 0.11876419931650162,
|
| 171 |
+
"ece": 0.022739037255446105,
|
| 172 |
+
"aurc": 0.016019787677674244,
|
| 173 |
+
"acc_at_80": 0.97,
|
| 174 |
"acc_at_50": 0.992,
|
| 175 |
"chance": 0.25,
|
| 176 |
+
"mean_conf": 0.9101096987724304,
|
| 177 |
+
"sec": 3.4,
|
| 178 |
"heldout": true
|
| 179 |
},
|
| 180 |
"dolly_category": {
|
| 181 |
"n": 1500,
|
| 182 |
+
"acc": 0.33266666666666667,
|
| 183 |
+
"nll": 2.009742021560669,
|
| 184 |
+
"brier": 0.8325088620185852,
|
| 185 |
+
"ece": 0.22487777996063235,
|
| 186 |
+
"aurc": 0.47846646274987925,
|
| 187 |
+
"acc_at_80": 0.3725,
|
| 188 |
+
"acc_at_50": 0.4493333333333333,
|
| 189 |
"chance": 0.125,
|
| 190 |
+
"mean_conf": 0.5575444102287292,
|
| 191 |
+
"sec": 3.8,
|
| 192 |
"heldout": true
|
| 193 |
},
|
| 194 |
"imdb": {
|
| 195 |
"n": 1500,
|
| 196 |
+
"acc": 0.9613333333333334,
|
| 197 |
+
"nll": 0.09619174897670746,
|
| 198 |
+
"brier": 0.057166311889886856,
|
| 199 |
+
"ece": 0.010577520211537642,
|
| 200 |
+
"aurc": 0.002844706294540048,
|
| 201 |
"acc_at_80": 0.9983333333333333,
|
| 202 |
"acc_at_50": 1.0,
|
| 203 |
"chance": 0.5,
|
| 204 |
+
"mean_conf": 0.9708292484283447,
|
| 205 |
+
"sec": 5.5,
|
| 206 |
"heldout": false
|
| 207 |
},
|
| 208 |
"sst2": {
|
| 209 |
"n": 872,
|
| 210 |
+
"acc": 0.9426605504587156,
|
| 211 |
+
"nll": 0.15219736099243164,
|
| 212 |
+
"brier": 0.08451657742261887,
|
| 213 |
+
"ece": 0.018561122545955384,
|
| 214 |
+
"aurc": 0.0094376600127498,
|
| 215 |
"acc_at_80": 0.9899713467048711,
|
| 216 |
"acc_at_50": 0.9931192660550459,
|
| 217 |
"chance": 0.5,
|
| 218 |
+
"mean_conf": 0.9517649412155151,
|
| 219 |
+
"sec": 3.8,
|
| 220 |
"heldout": false
|
| 221 |
},
|
| 222 |
"sst5": {
|
| 223 |
"n": 1500,
|
| 224 |
+
"acc": 0.6073333333333333,
|
| 225 |
+
"nll": 0.9326017498970032,
|
| 226 |
+
"brier": 0.5300925970077515,
|
| 227 |
+
"ece": 0.02625397855043411,
|
| 228 |
+
"aurc": 0.3098600910602178,
|
| 229 |
"acc_at_80": 0.6466666666666666,
|
| 230 |
+
"acc_at_50": 0.68,
|
| 231 |
"chance": 0.2,
|
| 232 |
+
"mean_conf": 0.5991732478141785,
|
| 233 |
+
"sec": 2.6,
|
| 234 |
"heldout": false
|
| 235 |
},
|
| 236 |
"yelp": {
|
| 237 |
"n": 1500,
|
| 238 |
+
"acc": 0.688,
|
| 239 |
+
"nll": 0.7123478055000305,
|
| 240 |
+
"brier": 0.41503453254699707,
|
| 241 |
+
"ece": 0.030766211390495284,
|
| 242 |
+
"aurc": 0.16381766100924522,
|
| 243 |
+
"acc_at_80": 0.7466666666666667,
|
| 244 |
"acc_at_50": 0.8146666666666667,
|
| 245 |
"chance": 0.2,
|
| 246 |
+
"mean_conf": 0.7086488604545593,
|
| 247 |
+
"sec": 4.6,
|
| 248 |
"heldout": false
|
| 249 |
},
|
| 250 |
"amazon_stars": {
|
| 251 |
"n": 1500,
|
| 252 |
+
"acc": 0.6073333333333333,
|
| 253 |
+
"nll": 0.9104115962982178,
|
| 254 |
+
"brier": 0.508173942565918,
|
| 255 |
+
"ece": 0.024688152035077397,
|
| 256 |
+
"aurc": 0.23917450560661985,
|
| 257 |
+
"acc_at_80": 0.6625,
|
| 258 |
+
"acc_at_50": 0.748,
|
| 259 |
"chance": 0.2,
|
| 260 |
+
"mean_conf": 0.6209068894386292,
|
| 261 |
+
"sec": 3.1,
|
| 262 |
"heldout": false
|
| 263 |
},
|
| 264 |
"emotion": {
|
| 265 |
"n": 1500,
|
| 266 |
+
"acc": 0.8526666666666667,
|
| 267 |
+
"nll": 0.3817766606807709,
|
| 268 |
+
"brier": 0.20716996490955353,
|
| 269 |
+
"ece": 0.028686144828796382,
|
| 270 |
+
"aurc": 0.036016887247968345,
|
| 271 |
+
"acc_at_80": 0.9241666666666667,
|
| 272 |
"acc_at_50": 0.9813333333333333,
|
| 273 |
"chance": 0.16666666666666666,
|
| 274 |
+
"mean_conf": 0.8751572370529175,
|
| 275 |
+
"sec": 2.6,
|
| 276 |
"heldout": false
|
| 277 |
},
|
| 278 |
"go_emotions": {
|
| 279 |
"n": 1500,
|
| 280 |
+
"acc": 0.77,
|
| 281 |
+
"nll": 0.6605949997901917,
|
| 282 |
+
"brier": 0.3219051659107208,
|
| 283 |
+
"ece": 0.02531181337436041,
|
| 284 |
+
"aurc": 0.07909983628498353,
|
| 285 |
+
"acc_at_80": 0.8525,
|
| 286 |
+
"acc_at_50": 0.9413333333333334,
|
| 287 |
"chance": 0.1,
|
| 288 |
+
"mean_conf": 0.7867252826690674,
|
| 289 |
+
"sec": 3.0,
|
| 290 |
"heldout": false
|
| 291 |
},
|
| 292 |
"tweet_sentiment": {
|
| 293 |
"n": 1500,
|
| 294 |
+
"acc": 0.732,
|
| 295 |
+
"nll": 0.5844112038612366,
|
| 296 |
+
"brier": 0.35898950695991516,
|
| 297 |
+
"ece": 0.023715959330399827,
|
| 298 |
+
"aurc": 0.1277814283266402,
|
| 299 |
+
"acc_at_80": 0.7908333333333334,
|
| 300 |
"acc_at_50": 0.868,
|
| 301 |
"chance": 0.3333333333333333,
|
| 302 |
+
"mean_conf": 0.7393827438354492,
|
| 303 |
"sec": 2.4,
|
| 304 |
"heldout": false
|
| 305 |
},
|
| 306 |
"tweet_emotion": {
|
| 307 |
"n": 1421,
|
| 308 |
+
"acc": 0.8332160450387052,
|
| 309 |
+
"nll": 0.4378463625907898,
|
| 310 |
+
"brier": 0.23256292939186096,
|
| 311 |
+
"ece": 0.029261003112725864,
|
| 312 |
+
"aurc": 0.0408950375447702,
|
| 313 |
+
"acc_at_80": 0.9129287598944591,
|
| 314 |
+
"acc_at_50": 0.9845070422535211,
|
| 315 |
"chance": 0.25,
|
| 316 |
+
"mean_conf": 0.8131034970283508,
|
| 317 |
+
"sec": 4.8,
|
| 318 |
"heldout": false
|
| 319 |
},
|
| 320 |
"tweet_irony": {
|
| 321 |
"n": 784,
|
| 322 |
+
"acc": 0.7933673469387755,
|
| 323 |
+
"nll": 0.4786851406097412,
|
| 324 |
+
"brier": 0.3143369257450104,
|
| 325 |
+
"ece": 0.05477343446442059,
|
| 326 |
+
"aurc": 0.11260140253357594,
|
| 327 |
+
"acc_at_80": 0.8261562998405104,
|
| 328 |
+
"acc_at_50": 0.8826530612244898,
|
| 329 |
"chance": 0.5,
|
| 330 |
+
"mean_conf": 0.7504966855049133,
|
| 331 |
+
"sec": 1.2,
|
| 332 |
"heldout": true
|
| 333 |
},
|
| 334 |
"fin_sentiment": {
|
| 335 |
"n": 1500,
|
| 336 |
+
"acc": 0.784,
|
| 337 |
+
"nll": 0.4931422472000122,
|
| 338 |
+
"brier": 0.29916250705718994,
|
| 339 |
+
"ece": 0.02881191293398539,
|
| 340 |
+
"aurc": 0.09077011331787095,
|
| 341 |
+
"acc_at_80": 0.8375,
|
| 342 |
+
"acc_at_50": 0.9266666666666666,
|
| 343 |
"chance": 0.3333333333333333,
|
| 344 |
+
"mean_conf": 0.7918360233306885,
|
| 345 |
+
"sec": 2.5,
|
| 346 |
"heldout": true
|
| 347 |
},
|
| 348 |
"cr_reviews": {
|
| 349 |
"n": 753,
|
| 350 |
+
"acc": 0.9070385126162018,
|
| 351 |
+
"nll": 0.23999790847301483,
|
| 352 |
+
"brier": 0.13726605474948883,
|
| 353 |
+
"ece": 0.02803241042976838,
|
| 354 |
+
"aurc": 0.019896790015697566,
|
| 355 |
+
"acc_at_80": 0.9617940199335548,
|
| 356 |
+
"acc_at_50": 0.9920212765957447,
|
| 357 |
"chance": 0.5,
|
| 358 |
+
"mean_conf": 0.9328383803367615,
|
| 359 |
+
"sec": 1.2,
|
| 360 |
"heldout": true
|
| 361 |
},
|
| 362 |
"counterfactual": {
|
| 363 |
"n": 1500,
|
| 364 |
+
"acc": 0.9546666666666667,
|
| 365 |
+
"nll": 0.1323074996471405,
|
| 366 |
+
"brier": 0.07490427792072296,
|
| 367 |
+
"ece": 0.014571970144907676,
|
| 368 |
+
"aurc": 0.0058697408516071295,
|
| 369 |
+
"acc_at_80": 0.99,
|
| 370 |
+
"acc_at_50": 0.9973333333333333,
|
| 371 |
"chance": 0.5,
|
| 372 |
+
"mean_conf": 0.9637404084205627,
|
| 373 |
+
"sec": 2.4,
|
| 374 |
"heldout": false
|
| 375 |
},
|
| 376 |
"subjectivity": {
|
| 377 |
"n": 1500,
|
| 378 |
+
"acc": 0.9633333333333334,
|
| 379 |
+
"nll": 0.11715030670166016,
|
| 380 |
+
"brier": 0.06305266171693802,
|
| 381 |
+
"ece": 0.011594313383102427,
|
| 382 |
+
"aurc": 0.005805298576335658,
|
| 383 |
"acc_at_80": 0.9916666666666667,
|
| 384 |
"acc_at_50": 0.9973333333333333,
|
| 385 |
"chance": 0.5,
|
| 386 |
+
"mean_conf": 0.9517436623573303,
|
| 387 |
"sec": 2.4,
|
| 388 |
"heldout": false
|
| 389 |
},
|
| 390 |
"tweet_offensive": {
|
| 391 |
"n": 860,
|
| 392 |
+
"acc": 0.8604651162790697,
|
| 393 |
+
"nll": 0.3307216167449951,
|
| 394 |
+
"brier": 0.2062027007341385,
|
| 395 |
+
"ece": 0.03277220109174419,
|
| 396 |
+
"aurc": 0.043970018789679495,
|
| 397 |
+
"acc_at_80": 0.9186046511627907,
|
| 398 |
+
"acc_at_50": 0.9627906976744186,
|
| 399 |
"chance": 0.5,
|
| 400 |
+
"mean_conf": 0.8332623243331909,
|
| 401 |
+
"sec": 1.5,
|
| 402 |
"heldout": false
|
| 403 |
},
|
| 404 |
"tweet_hate": {
|
| 405 |
"n": 1500,
|
| 406 |
+
"acc": 0.5106666666666667,
|
| 407 |
+
"nll": 0.951435387134552,
|
| 408 |
+
"brier": 0.6554798483848572,
|
| 409 |
+
"ece": 0.27647322726249696,
|
| 410 |
+
"aurc": 0.421070973829564,
|
| 411 |
"acc_at_80": 0.535,
|
| 412 |
+
"acc_at_50": 0.556,
|
| 413 |
"chance": 0.5,
|
| 414 |
+
"mean_conf": 0.7855966687202454,
|
| 415 |
+
"sec": 2.6,
|
| 416 |
"heldout": false
|
| 417 |
},
|
| 418 |
"hate_offensive": {
|
| 419 |
"n": 1500,
|
| 420 |
+
"acc": 0.9206666666666666,
|
| 421 |
+
"nll": 0.2384883612394333,
|
| 422 |
+
"brier": 0.1258707195520401,
|
| 423 |
+
"ece": 0.013242818454901406,
|
| 424 |
+
"aurc": 0.018655782825341898,
|
| 425 |
"acc_at_80": 0.9716666666666667,
|
| 426 |
+
"acc_at_50": 0.9853333333333333,
|
| 427 |
"chance": 0.3333333333333333,
|
| 428 |
+
"mean_conf": 0.9204931855201721,
|
| 429 |
+
"sec": 2.6,
|
| 430 |
"heldout": false
|
| 431 |
},
|
| 432 |
"civil_comments": {
|
| 433 |
"n": 7500,
|
| 434 |
+
"acc": 0.9356,
|
| 435 |
+
"nll": 0.15526148676872253,
|
| 436 |
+
"brier": 0.09336414933204651,
|
| 437 |
+
"ece": 0.013219891810417186,
|
| 438 |
+
"aurc": 0.007861904760099968,
|
| 439 |
+
"acc_at_80": 0.9856666666666667,
|
| 440 |
"acc_at_50": 0.9997333333333334,
|
| 441 |
"chance": 0.5,
|
| 442 |
+
"mean_conf": 0.9488199353218079,
|
| 443 |
+
"sec": 4.4,
|
| 444 |
"heldout": false,
|
| 445 |
"per_q": [
|
| 446 |
+
0.8753333333333333,
|
| 447 |
+
0.9786666666666667,
|
| 448 |
+
0.9913333333333333,
|
| 449 |
+
0.8633333333333333,
|
| 450 |
0.9693333333333334
|
| 451 |
]
|
| 452 |
},
|
| 453 |
"toxic_chat": {
|
| 454 |
"n": 3000,
|
| 455 |
+
"acc": 0.9793333333333333,
|
| 456 |
+
"nll": 0.05689281225204468,
|
| 457 |
+
"brier": 0.029941946268081665,
|
| 458 |
+
"ece": 0.008714853028456338,
|
| 459 |
+
"aurc": 0.00120893581584804,
|
| 460 |
+
"acc_at_80": 0.9991666666666666,
|
| 461 |
+
"acc_at_50": 0.9993333333333333,
|
| 462 |
"chance": 0.5,
|
| 463 |
+
"mean_conf": 0.985617458820343,
|
| 464 |
+
"sec": 3.3,
|
| 465 |
"heldout": false,
|
| 466 |
"per_q": [
|
| 467 |
+
0.964,
|
| 468 |
+
0.9946666666666667
|
| 469 |
]
|
| 470 |
},
|
| 471 |
"sms_spam": {
|
| 472 |
"n": 1000,
|
| 473 |
+
"acc": 0.984,
|
| 474 |
+
"nll": 0.054738420993089676,
|
| 475 |
+
"brier": 0.023743100464344025,
|
| 476 |
+
"ece": 0.009705231130123096,
|
| 477 |
+
"aurc": 0.002162345807605099,
|
| 478 |
"acc_at_80": 0.99625,
|
| 479 |
"acc_at_50": 0.998,
|
| 480 |
"chance": 0.5,
|
| 481 |
+
"mean_conf": 0.9895715117454529,
|
| 482 |
+
"sec": 4.2,
|
| 483 |
"heldout": false
|
| 484 |
},
|
| 485 |
"enron_spam": {
|
| 486 |
"n": 1500,
|
| 487 |
+
"acc": 0.9866666666666667,
|
| 488 |
+
"nll": 0.02601483091711998,
|
| 489 |
+
"brier": 0.016239631921052933,
|
| 490 |
+
"ece": 0.004016611814498852,
|
| 491 |
+
"aurc": 0.0002021199142868427,
|
| 492 |
"acc_at_80": 1.0,
|
| 493 |
"acc_at_50": 1.0,
|
| 494 |
"chance": 0.5,
|
| 495 |
+
"mean_conf": 0.9905825257301331,
|
| 496 |
+
"sec": 9.2,
|
| 497 |
"heldout": false
|
| 498 |
},
|
| 499 |
"insincere_questions": {
|
| 500 |
"n": 1500,
|
| 501 |
"acc": 0.9553333333333334,
|
| 502 |
+
"nll": 0.10723885893821716,
|
| 503 |
+
"brier": 0.06371062248945236,
|
| 504 |
+
"ece": 0.011764580051104229,
|
| 505 |
+
"aurc": 0.004172164123818539,
|
| 506 |
"acc_at_80": 0.9975,
|
| 507 |
"acc_at_50": 0.9986666666666667,
|
| 508 |
"chance": 0.5,
|
| 509 |
+
"mean_conf": 0.9626001119613647,
|
| 510 |
"sec": 2.4,
|
| 511 |
"heldout": false
|
| 512 |
},
|
| 513 |
"ade": {
|
| 514 |
"n": 1500,
|
| 515 |
+
"acc": 0.7893333333333333,
|
| 516 |
+
"nll": 0.4881707429885864,
|
| 517 |
+
"brier": 0.3113546073436737,
|
| 518 |
+
"ece": 0.0730705860455831,
|
| 519 |
+
"aurc": 0.10379213741472014,
|
| 520 |
+
"acc_at_80": 0.8325,
|
| 521 |
+
"acc_at_50": 0.896,
|
| 522 |
"chance": 0.5,
|
| 523 |
+
"mean_conf": 0.8475932478904724,
|
| 524 |
+
"sec": 2.5,
|
| 525 |
"heldout": true
|
| 526 |
},
|
| 527 |
"snli": {
|
| 528 |
"n": 1476,
|
| 529 |
+
"acc": 0.9234417344173442,
|
| 530 |
+
"nll": 0.22889889776706696,
|
| 531 |
+
"brier": 0.1168968454003334,
|
| 532 |
+
"ece": 0.011546386673360035,
|
| 533 |
+
"aurc": 0.017872040290664873,
|
| 534 |
"acc_at_80": 0.9729043183742591,
|
| 535 |
+
"acc_at_50": 0.9850948509485095,
|
| 536 |
"chance": 0.3333333333333333,
|
| 537 |
+
"mean_conf": 0.9265872240066528,
|
| 538 |
+
"sec": 11.6,
|
| 539 |
"heldout": false
|
| 540 |
},
|
| 541 |
"mnli": {
|
| 542 |
"n": 1500,
|
| 543 |
+
"acc": 0.878,
|
| 544 |
+
"nll": 0.29665282368659973,
|
| 545 |
+
"brier": 0.16776761412620544,
|
| 546 |
+
"ece": 0.028746872107187925,
|
| 547 |
+
"aurc": 0.024781012578722568,
|
| 548 |
+
"acc_at_80": 0.9558333333333333,
|
| 549 |
+
"acc_at_50": 0.9893333333333333,
|
| 550 |
"chance": 0.3333333333333333,
|
| 551 |
+
"mean_conf": 0.8953821659088135,
|
| 552 |
+
"sec": 3.1,
|
| 553 |
"heldout": false
|
| 554 |
},
|
| 555 |
"rte": {
|
| 556 |
"n": 277,
|
| 557 |
+
"acc": 0.8736462093862816,
|
| 558 |
+
"nll": 0.3355311453342438,
|
| 559 |
+
"brier": 0.20029526948928833,
|
| 560 |
+
"ece": 0.049293557659383265,
|
| 561 |
+
"aurc": 0.04237874308319795,
|
| 562 |
+
"acc_at_80": 0.9144144144144144,
|
| 563 |
+
"acc_at_50": 0.9637681159420289,
|
| 564 |
"chance": 0.5,
|
| 565 |
+
"mean_conf": 0.9088274240493774,
|
| 566 |
+
"sec": 0.7,
|
| 567 |
"heldout": false
|
| 568 |
},
|
| 569 |
"qnli": {
|
| 570 |
"n": 1500,
|
| 571 |
+
"acc": 0.9386666666666666,
|
| 572 |
+
"nll": 0.15850575268268585,
|
| 573 |
+
"brier": 0.0912380963563919,
|
| 574 |
+
"ece": 0.008112372795740749,
|
| 575 |
+
"aurc": 0.009528591919053303,
|
| 576 |
"acc_at_80": 0.9841666666666666,
|
| 577 |
+
"acc_at_50": 0.996,
|
| 578 |
"chance": 0.5,
|
| 579 |
+
"mean_conf": 0.9442304968833923,
|
| 580 |
+
"sec": 3.0,
|
| 581 |
"heldout": false
|
| 582 |
},
|
| 583 |
"qqp": {
|
| 584 |
"n": 1500,
|
| 585 |
+
"acc": 0.8546666666666667,
|
| 586 |
+
"nll": 0.3271322548389435,
|
| 587 |
+
"brier": 0.20945638418197632,
|
| 588 |
+
"ece": 0.026240944465001415,
|
| 589 |
+
"aurc": 0.04136483929310011,
|
| 590 |
+
"acc_at_80": 0.9116666666666666,
|
| 591 |
+
"acc_at_50": 0.976,
|
| 592 |
"chance": 0.5,
|
| 593 |
+
"mean_conf": 0.8754271864891052,
|
| 594 |
+
"sec": 2.6,
|
| 595 |
"heldout": false
|
| 596 |
},
|
| 597 |
"mrpc": {
|
| 598 |
"n": 408,
|
| 599 |
"acc": 0.8455882352941176,
|
| 600 |
+
"nll": 0.33147454261779785,
|
| 601 |
+
"brier": 0.2129886895418167,
|
| 602 |
+
"ece": 0.053440071058039586,
|
| 603 |
+
"aurc": 0.04242757999322468,
|
| 604 |
+
"acc_at_80": 0.9141104294478528,
|
| 605 |
"acc_at_50": 0.9803921568627451,
|
| 606 |
"chance": 0.5,
|
| 607 |
+
"mean_conf": 0.8646318912506104,
|
| 608 |
+
"sec": 0.8,
|
| 609 |
"heldout": false
|
| 610 |
},
|
| 611 |
"paws": {
|
| 612 |
"n": 1500,
|
| 613 |
+
"acc": 0.7013333333333334,
|
| 614 |
+
"nll": 0.8306891918182373,
|
| 615 |
+
"brier": 0.4740028381347656,
|
| 616 |
+
"ece": 0.22124919974803925,
|
| 617 |
+
"aurc": 0.13745002888739216,
|
| 618 |
+
"acc_at_80": 0.77,
|
| 619 |
+
"acc_at_50": 0.888,
|
| 620 |
"chance": 0.5,
|
| 621 |
+
"mean_conf": 0.9219720959663391,
|
| 622 |
+
"sec": 3.0,
|
| 623 |
"heldout": true
|
| 624 |
},
|
| 625 |
"cola": {
|
| 626 |
"n": 1043,
|
| 627 |
+
"acc": 0.8024928092042186,
|
| 628 |
+
"nll": 0.4121968150138855,
|
| 629 |
+
"brier": 0.26549777388572693,
|
| 630 |
+
"ece": 0.03725165488735943,
|
| 631 |
+
"aurc": 0.07404416991022049,
|
| 632 |
+
"acc_at_80": 0.8741007194244604,
|
| 633 |
"acc_at_50": 0.9367816091954023,
|
| 634 |
"chance": 0.5,
|
| 635 |
+
"mean_conf": 0.8369820713996887,
|
| 636 |
+
"sec": 1.6,
|
| 637 |
"heldout": false
|
| 638 |
},
|
| 639 |
"boolq": {
|
| 640 |
"n": 1500,
|
| 641 |
+
"acc": 0.8853333333333333,
|
| 642 |
+
"nll": 0.2897628843784332,
|
| 643 |
+
"brier": 0.17195458710193634,
|
| 644 |
+
"ece": 0.028059083223342877,
|
| 645 |
+
"aurc": 0.033273033407070704,
|
| 646 |
+
"acc_at_80": 0.9441666666666667,
|
| 647 |
+
"acc_at_50": 0.9746666666666667,
|
| 648 |
"chance": 0.5,
|
| 649 |
+
"mean_conf": 0.9083826541900635,
|
| 650 |
+
"sec": 4.0,
|
| 651 |
"heldout": false
|
| 652 |
},
|
| 653 |
"strategyqa": {
|
| 654 |
"n": 687,
|
| 655 |
+
"acc": 0.6142649199417758,
|
| 656 |
+
"nll": 0.6972374320030212,
|
| 657 |
+
"brier": 0.4870849847793579,
|
| 658 |
+
"ece": 0.10019853498564363,
|
| 659 |
+
"aurc": 0.31452270422741296,
|
| 660 |
+
"acc_at_80": 0.6327272727272727,
|
| 661 |
+
"acc_at_50": 0.6569767441860465,
|
| 662 |
"chance": 0.5,
|
| 663 |
+
"mean_conf": 0.7026323080062866,
|
| 664 |
+
"sec": 7.9,
|
| 665 |
"heldout": true
|
| 666 |
},
|
| 667 |
"pubmedqa": {
|
| 668 |
"n": 500,
|
| 669 |
+
"acc": 0.798,
|
| 670 |
+
"nll": 0.5936447381973267,
|
| 671 |
+
"brier": 0.31999242305755615,
|
| 672 |
+
"ece": 0.06877634388208391,
|
| 673 |
+
"aurc": 0.10006724050038115,
|
| 674 |
+
"acc_at_80": 0.845,
|
| 675 |
+
"acc_at_50": 0.928,
|
| 676 |
"chance": 0.33333333333333326,
|
| 677 |
+
"mean_conf": 0.8425376415252686,
|
| 678 |
+
"sec": 2.3,
|
| 679 |
"heldout": true
|
| 680 |
},
|
| 681 |
"arc": {
|
| 682 |
"n": 1172,
|
| 683 |
+
"acc": 0.856655290102389,
|
| 684 |
+
"nll": 0.3848453760147095,
|
| 685 |
+
"brier": 0.20024119317531586,
|
| 686 |
+
"ece": 0.03268225236760882,
|
| 687 |
+
"aurc": 0.028683364797018013,
|
| 688 |
+
"acc_at_80": 0.9466950959488273,
|
| 689 |
+
"acc_at_50": 0.9897610921501706,
|
| 690 |
"chance": 0.25015642775881686,
|
| 691 |
+
"mean_conf": 0.8850091695785522,
|
| 692 |
+
"sec": 2.3,
|
| 693 |
"heldout": false
|
| 694 |
},
|
| 695 |
"commonsense_qa": {
|
| 696 |
"n": 1221,
|
| 697 |
+
"acc": 0.7641277641277642,
|
| 698 |
+
"nll": 0.6035410165786743,
|
| 699 |
+
"brier": 0.32016345858573914,
|
| 700 |
+
"ece": 0.024896302787521474,
|
| 701 |
+
"aurc": 0.07891618232675111,
|
| 702 |
+
"acc_at_80": 0.8454452405322416,
|
| 703 |
+
"acc_at_50": 0.9459016393442623,
|
| 704 |
"chance": 0.19999999999999998,
|
| 705 |
+
"mean_conf": 0.7828702926635742,
|
| 706 |
"sec": 2.0,
|
| 707 |
"heldout": false
|
| 708 |
},
|
| 709 |
"qasc": {
|
| 710 |
"n": 926,
|
| 711 |
+
"acc": 0.8142548596112311,
|
| 712 |
+
"nll": 0.5944124460220337,
|
| 713 |
+
"brier": 0.27712827920913696,
|
| 714 |
+
"ece": 0.05084534546211523,
|
| 715 |
+
"aurc": 0.04955932347383356,
|
| 716 |
+
"acc_at_80": 0.8960863697705803,
|
| 717 |
+
"acc_at_50": 0.9676025917926566,
|
| 718 |
"chance": 0.125,
|
| 719 |
+
"mean_conf": 0.7678154706954956,
|
| 720 |
+
"sec": 1.8,
|
| 721 |
"heldout": false
|
| 722 |
},
|
| 723 |
"openbookqa": {
|
| 724 |
"n": 500,
|
| 725 |
+
"acc": 0.808,
|
| 726 |
+
"nll": 0.5015584826469421,
|
| 727 |
+
"brier": 0.25752735137939453,
|
| 728 |
+
"ece": 0.044371710240840904,
|
| 729 |
+
"aurc": 0.057444096145339726,
|
| 730 |
"acc_at_80": 0.905,
|
| 731 |
+
"acc_at_50": 0.968,
|
| 732 |
"chance": 0.25,
|
| 733 |
+
"mean_conf": 0.8454900979995728,
|
| 734 |
+
"sec": 0.9,
|
| 735 |
"heldout": false
|
| 736 |
},
|
| 737 |
"sciq": {
|
| 738 |
"n": 1000,
|
| 739 |
+
"acc": 0.983,
|
| 740 |
+
"nll": 0.0540383905172348,
|
| 741 |
+
"brier": 0.026997487992048264,
|
| 742 |
+
"ece": 0.015873215794563272,
|
| 743 |
+
"aurc": 0.0004725583644347805,
|
| 744 |
"acc_at_80": 1.0,
|
| 745 |
"acc_at_50": 1.0,
|
| 746 |
"chance": 0.25,
|
| 747 |
+
"mean_conf": 0.9763537645339966,
|
| 748 |
+
"sec": 4.9,
|
| 749 |
"heldout": true
|
| 750 |
},
|
| 751 |
"hellaswag": {
|
| 752 |
"n": 1500,
|
| 753 |
+
"acc": 0.8726666666666667,
|
| 754 |
+
"nll": 0.35282617807388306,
|
| 755 |
+
"brier": 0.1844407618045807,
|
| 756 |
+
"ece": 0.05096171965201693,
|
| 757 |
+
"aurc": 0.022184705622736417,
|
| 758 |
+
"acc_at_80": 0.9541666666666667,
|
| 759 |
"acc_at_50": 0.9986666666666667,
|
| 760 |
"chance": 0.25,
|
| 761 |
+
"mean_conf": 0.9218340516090393,
|
| 762 |
+
"sec": 4.4,
|
| 763 |
"heldout": false
|
| 764 |
},
|
| 765 |
"piqa": {
|
| 766 |
"n": 1500,
|
| 767 |
+
"acc": 0.8353333333333334,
|
| 768 |
+
"nll": 0.3836018443107605,
|
| 769 |
+
"brier": 0.2365727722644806,
|
| 770 |
+
"ece": 0.04386041740576427,
|
| 771 |
+
"aurc": 0.05569590131269642,
|
| 772 |
+
"acc_at_80": 0.8966666666666666,
|
| 773 |
+
"acc_at_50": 0.9506666666666667,
|
| 774 |
"chance": 0.5,
|
| 775 |
+
"mean_conf": 0.8791936635971069,
|
| 776 |
+
"sec": 2.9,
|
| 777 |
"heldout": false
|
| 778 |
},
|
| 779 |
"social_iqa": {
|
| 780 |
"n": 1500,
|
| 781 |
+
"acc": 0.7186666666666667,
|
| 782 |
+
"nll": 0.7424599528312683,
|
| 783 |
+
"brier": 0.41328999400138855,
|
| 784 |
+
"ece": 0.09968751738468809,
|
| 785 |
+
"aurc": 0.1436145844762169,
|
| 786 |
+
"acc_at_80": 0.775,
|
| 787 |
+
"acc_at_50": 0.8546666666666667,
|
| 788 |
"chance": 0.3333333333333333,
|
| 789 |
+
"mean_conf": 0.8163115382194519,
|
| 790 |
+
"sec": 2.7,
|
| 791 |
"heldout": true
|
| 792 |
},
|
| 793 |
"winogrande": {
|
| 794 |
"n": 1267,
|
| 795 |
+
"acc": 0.681136543014996,
|
| 796 |
+
"nll": 0.7226526737213135,
|
| 797 |
+
"brier": 0.45336562395095825,
|
| 798 |
+
"ece": 0.16130592141832623,
|
| 799 |
+
"aurc": 0.19273822608019725,
|
| 800 |
+
"acc_at_80": 0.7209072978303748,
|
| 801 |
+
"acc_at_50": 0.8123028391167192,
|
| 802 |
"chance": 0.5,
|
| 803 |
+
"mean_conf": 0.8380995988845825,
|
| 804 |
+
"sec": 1.9,
|
| 805 |
"heldout": false
|
| 806 |
},
|
| 807 |
"race": {
|
| 808 |
"n": 1500,
|
| 809 |
+
"acc": 0.8326666666666667,
|
| 810 |
+
"nll": 0.48354849219322205,
|
| 811 |
+
"brier": 0.24252651631832123,
|
| 812 |
+
"ece": 0.05754320712884267,
|
| 813 |
+
"aurc": 0.04422304284234363,
|
| 814 |
+
"acc_at_80": 0.9175,
|
| 815 |
+
"acc_at_50": 0.9733333333333334,
|
| 816 |
"chance": 0.25,
|
| 817 |
+
"mean_conf": 0.8896796703338623,
|
| 818 |
+
"sec": 7.3,
|
| 819 |
"heldout": false
|
| 820 |
},
|
| 821 |
"mmlu": {
|
| 822 |
"n": 1500,
|
| 823 |
+
"acc": 0.6333333333333333,
|
| 824 |
+
"nll": 0.9389562010765076,
|
| 825 |
+
"brier": 0.47845685482025146,
|
| 826 |
+
"ece": 0.10005126521984736,
|
| 827 |
+
"aurc": 0.16047564173919576,
|
| 828 |
+
"acc_at_80": 0.71,
|
| 829 |
+
"acc_at_50": 0.8533333333333334,
|
| 830 |
"chance": 0.25,
|
| 831 |
+
"mean_conf": 0.7325466871261597,
|
| 832 |
+
"sec": 3.4,
|
| 833 |
"heldout": false
|
| 834 |
},
|
| 835 |
"medqa": {
|
| 836 |
"n": 1273,
|
| 837 |
+
"acc": 0.5632364493322859,
|
| 838 |
+
"nll": 1.079268455505371,
|
| 839 |
+
"brier": 0.575121283531189,
|
| 840 |
+
"ece": 0.1046450646112776,
|
| 841 |
+
"aurc": 0.2684613401796164,
|
| 842 |
+
"acc_at_80": 0.6129666011787819,
|
| 843 |
+
"acc_at_50": 0.710691823899371,
|
| 844 |
"chance": 0.25,
|
| 845 |
+
"mean_conf": 0.6660441756248474,
|
| 846 |
+
"sec": 4.1,
|
| 847 |
"heldout": false
|
| 848 |
},
|
| 849 |
"truthfulqa": {
|
| 850 |
"n": 817,
|
| 851 |
+
"acc": 0.5018359853121175,
|
| 852 |
+
"nll": 1.5319370031356812,
|
| 853 |
+
"brier": 0.6671698689460754,
|
| 854 |
+
"ece": 0.1319449910919115,
|
| 855 |
+
"aurc": 0.2963526198417848,
|
| 856 |
+
"acc_at_80": 0.5581039755351682,
|
| 857 |
+
"acc_at_50": 0.678921568627451,
|
| 858 |
"chance": 0.22622350449767833,
|
| 859 |
+
"mean_conf": 0.6299614310264587,
|
| 860 |
+
"sec": 1.7,
|
| 861 |
"heldout": true
|
| 862 |
},
|
| 863 |
"fin_phrasebank": {
|
| 864 |
"n": 970,
|
| 865 |
+
"acc": 0.6989690721649484,
|
| 866 |
+
"nll": 0.6254894733428955,
|
| 867 |
+
"brier": 0.39401963353157043,
|
| 868 |
+
"ece": 0.058691233703770586,
|
| 869 |
+
"aurc": 0.16686274774956414,
|
| 870 |
+
"acc_at_80": 0.7577319587628866,
|
| 871 |
+
"acc_at_50": 0.8329896907216495,
|
| 872 |
"chance": 0.33333333333333326,
|
| 873 |
+
"mean_conf": 0.7559286952018738,
|
| 874 |
+
"sec": 6.3,
|
| 875 |
"heldout": true
|
| 876 |
},
|
| 877 |
"banking77": {
|
| 878 |
"n": 1500,
|
| 879 |
+
"acc": 0.9626666666666667,
|
| 880 |
+
"nll": 0.12322639673948288,
|
| 881 |
+
"brier": 0.05743316188454628,
|
| 882 |
+
"ece": 0.007931326011816654,
|
| 883 |
+
"aurc": 0.002740366914790048,
|
| 884 |
"acc_at_80": 0.9966666666666667,
|
| 885 |
"acc_at_50": 1.0,
|
| 886 |
"chance": 0.1,
|
| 887 |
+
"mean_conf": 0.9672797918319702,
|
| 888 |
+
"sec": 3.0,
|
| 889 |
"heldout": false
|
| 890 |
},
|
| 891 |
"bias_in_bios": {
|
| 892 |
"n": 3000,
|
| 893 |
+
"acc": 0.9516666666666667,
|
| 894 |
+
"nll": 0.13784465193748474,
|
| 895 |
+
"brier": 0.06864367425441742,
|
| 896 |
+
"ece": 0.006546956946452431,
|
| 897 |
+
"aurc": 0.0033626200148913327,
|
| 898 |
+
"acc_at_80": 0.9975,
|
| 899 |
"acc_at_50": 1.0,
|
| 900 |
"chance": 0.30000000000000004,
|
| 901 |
+
"mean_conf": 0.953985333442688,
|
| 902 |
+
"sec": 4.2,
|
| 903 |
"heldout": false,
|
| 904 |
"per_q": [
|
| 905 |
+
0.906,
|
| 906 |
0.9973333333333333
|
| 907 |
]
|
| 908 |
},
|
| 909 |
"helpsteer2": {
|
| 910 |
"n": 5190,
|
| 911 |
+
"acc": 0.6028901734104046,
|
| 912 |
+
"nll": 0.9502481818199158,
|
| 913 |
+
"brier": 0.5227839350700378,
|
| 914 |
+
"ece": 0.02052123054201663,
|
| 915 |
+
"aurc": 0.2681509263336261,
|
| 916 |
+
"acc_at_80": 0.6623314065510597,
|
| 917 |
+
"acc_at_50": 0.7294797687861272,
|
| 918 |
"chance": 0.2,
|
| 919 |
+
"mean_conf": 0.6144348978996277,
|
| 920 |
+
"sec": 23.2,
|
| 921 |
"heldout": false,
|
| 922 |
"per_q": [
|
| 923 |
+
0.45664739884393063,
|
| 924 |
+
0.5067437379576107,
|
| 925 |
+
0.7225433526011561,
|
| 926 |
+
0.6358381502890174,
|
| 927 |
+
0.6926782273603083
|
| 928 |
]
|
| 929 |
},
|
| 930 |
"helpsteer3_pref": {
|
| 931 |
"n": 1176,
|
| 932 |
+
"acc": 0.4005102040816326,
|
| 933 |
+
"nll": 1.457674503326416,
|
| 934 |
+
"brier": 0.7135226726531982,
|
| 935 |
+
"ece": 0.05034625236274434,
|
| 936 |
+
"aurc": 0.49111859183201545,
|
| 937 |
+
"acc_at_80": 0.4293304994686504,
|
| 938 |
+
"acc_at_50": 0.49489795918367346,
|
| 939 |
"chance": 0.14285714285714282,
|
| 940 |
+
"mean_conf": 0.44308316707611084,
|
| 941 |
+
"sec": 24.6,
|
| 942 |
"heldout": false
|
| 943 |
},
|
| 944 |
"hate_speech_scales": {
|
| 945 |
"n": 7500,
|
| 946 |
+
"acc": 0.5605333333333333,
|
| 947 |
+
"nll": 0.9819027781486511,
|
| 948 |
+
"brier": 0.5431735515594482,
|
| 949 |
+
"ece": 0.0195502276579539,
|
| 950 |
+
"aurc": 0.267479078510345,
|
| 951 |
+
"acc_at_80": 0.6098333333333333,
|
| 952 |
+
"acc_at_50": 0.7061333333333333,
|
| 953 |
"chance": 0.22666666666666674,
|
| 954 |
+
"mean_conf": 0.5698475241661072,
|
| 955 |
+
"sec": 4.9,
|
| 956 |
"heldout": false,
|
| 957 |
"per_q": [
|
| 958 |
0.706,
|
| 959 |
+
0.4826666666666667,
|
| 960 |
+
0.43533333333333335,
|
| 961 |
+
0.578,
|
| 962 |
+
0.6006666666666667
|
| 963 |
]
|
| 964 |
},
|
| 965 |
"liar2": {
|
| 966 |
"n": 1500,
|
| 967 |
+
"acc": 0.37466666666666665,
|
| 968 |
+
"nll": 1.465606451034546,
|
| 969 |
+
"brier": 0.7157273888587952,
|
| 970 |
+
"ece": 0.028325774570306134,
|
| 971 |
+
"aurc": 0.47234860740182666,
|
| 972 |
+
"acc_at_80": 0.4075,
|
| 973 |
+
"acc_at_50": 0.48933333333333334,
|
| 974 |
"chance": 0.16666666666666666,
|
| 975 |
+
"mean_conf": 0.39819955825805664,
|
| 976 |
+
"sec": 3.7,
|
| 977 |
"heldout": false
|
| 978 |
},
|
| 979 |
"prosocial_safety": {
|
| 980 |
"n": 1500,
|
| 981 |
+
"acc": 0.524,
|
| 982 |
+
"nll": 1.1723508834838867,
|
| 983 |
+
"brier": 0.592208206653595,
|
| 984 |
+
"ece": 0.02242480476697286,
|
| 985 |
+
"aurc": 0.2745796786723848,
|
| 986 |
+
"acc_at_80": 0.5883333333333334,
|
| 987 |
+
"acc_at_50": 0.6906666666666667,
|
| 988 |
"chance": 0.2,
|
| 989 |
+
"mean_conf": 0.5037593841552734,
|
| 990 |
+
"sec": 3.0,
|
| 991 |
"heldout": false
|
| 992 |
},
|
| 993 |
"ultrafeedback_pref": {
|
| 994 |
"n": 1500,
|
| 995 |
+
"acc": 0.762,
|
| 996 |
+
"nll": 0.5141094923019409,
|
| 997 |
+
"brier": 0.33274728059768677,
|
| 998 |
+
"ece": 0.03840167435010275,
|
| 999 |
+
"aurc": 0.1266040181966966,
|
| 1000 |
+
"acc_at_80": 0.81,
|
| 1001 |
+
"acc_at_50": 0.864,
|
| 1002 |
"chance": 0.5,
|
| 1003 |
+
"mean_conf": 0.794404923915863,
|
| 1004 |
+
"sec": 9.8,
|
| 1005 |
"heldout": false
|
| 1006 |
},
|
| 1007 |
"shp": {
|
| 1008 |
"n": 1500,
|
| 1009 |
+
"acc": 0.744,
|
| 1010 |
+
"nll": 0.5253914594650269,
|
| 1011 |
+
"brier": 0.35134565830230713,
|
| 1012 |
+
"ece": 0.029734440008799243,
|
| 1013 |
+
"aurc": 0.14422793871032347,
|
| 1014 |
+
"acc_at_80": 0.7808333333333334,
|
| 1015 |
+
"acc_at_50": 0.848,
|
| 1016 |
"chance": 0.5,
|
| 1017 |
+
"mean_conf": 0.752452552318573,
|
| 1018 |
+
"sec": 7.2,
|
| 1019 |
"heldout": false
|
| 1020 |
},
|
| 1021 |
"hh_rlhf": {
|
| 1022 |
"n": 1488,
|
| 1023 |
+
"acc": 0.6478494623655914,
|
| 1024 |
+
"nll": 0.6344334483146667,
|
| 1025 |
+
"brier": 0.44163978099823,
|
| 1026 |
+
"ece": 0.06129897718307793,
|
| 1027 |
+
"aurc": 0.24894185193228557,
|
| 1028 |
+
"acc_at_80": 0.6705882352941176,
|
| 1029 |
+
"acc_at_50": 0.7338709677419355,
|
| 1030 |
"chance": 0.5,
|
| 1031 |
+
"mean_conf": 0.6988418698310852,
|
| 1032 |
+
"sec": 12.3,
|
| 1033 |
"heldout": false
|
| 1034 |
},
|
| 1035 |
"arena_pref": {
|
| 1036 |
"n": 1500,
|
| 1037 |
+
"acc": 0.47533333333333333,
|
| 1038 |
+
"nll": 1.2214066982269287,
|
| 1039 |
+
"brier": 0.6920773386955261,
|
| 1040 |
+
"ece": 0.1775780066649119,
|
| 1041 |
+
"aurc": 0.4264101430176788,
|
| 1042 |
+
"acc_at_80": 0.5025,
|
| 1043 |
+
"acc_at_50": 0.5493333333333333,
|
| 1044 |
"chance": 0.3333333333333333,
|
| 1045 |
+
"mean_conf": 0.652911365032196,
|
| 1046 |
+
"sec": 9.3,
|
| 1047 |
"heldout": true
|
| 1048 |
},
|
| 1049 |
"reward_bench": {
|
| 1050 |
"n": 1500,
|
| 1051 |
+
"acc": 0.794,
|
| 1052 |
+
"nll": 0.45838916301727295,
|
| 1053 |
+
"brier": 0.28988850116729736,
|
| 1054 |
+
"ece": 0.04195846394697825,
|
| 1055 |
+
"aurc": 0.09336704492407273,
|
| 1056 |
+
"acc_at_80": 0.8533333333333334,
|
| 1057 |
+
"acc_at_50": 0.912,
|
| 1058 |
"chance": 0.5,
|
| 1059 |
+
"mean_conf": 0.8290508389472961,
|
| 1060 |
+
"sec": 7.7,
|
| 1061 |
"heldout": true
|
| 1062 |
},
|
| 1063 |
"glaive_tools": {
|
| 1064 |
"n": 1348,
|
| 1065 |
+
"acc": 0.9473293768545994,
|
| 1066 |
+
"nll": 0.1643306165933609,
|
| 1067 |
+
"brier": 0.087461918592453,
|
| 1068 |
+
"ece": 0.007846767320477323,
|
| 1069 |
+
"aurc": 0.011936188106662745,
|
| 1070 |
+
"acc_at_80": 0.9795918367346939,
|
| 1071 |
+
"acc_at_50": 0.9925816023738873,
|
| 1072 |
"chance": 0.16666666666666666,
|
| 1073 |
+
"mean_conf": 0.9517403841018677,
|
| 1074 |
+
"sec": 9.1,
|
| 1075 |
"heldout": false
|
| 1076 |
},
|
| 1077 |
"hermes_tools": {
|
| 1078 |
"n": 1500,
|
| 1079 |
+
"acc": 0.7273333333333334,
|
| 1080 |
+
"nll": 0.5851268768310547,
|
| 1081 |
+
"brier": 0.3813258111476898,
|
| 1082 |
+
"ece": 0.19555236707131068,
|
| 1083 |
+
"aurc": 0.06225953895023327,
|
| 1084 |
+
"acc_at_80": 0.8558333333333333,
|
| 1085 |
+
"acc_at_50": 0.988,
|
| 1086 |
"chance": 0.125,
|
| 1087 |
+
"mean_conf": 0.92147296667099,
|
| 1088 |
+
"sec": 8.8,
|
| 1089 |
"heldout": true
|
| 1090 |
},
|
| 1091 |
"copa": {
|
| 1092 |
"n": 100,
|
| 1093 |
+
"acc": 0.94,
|
| 1094 |
+
"nll": 0.1961805373430252,
|
| 1095 |
+
"brier": 0.09868918359279633,
|
| 1096 |
+
"ece": 0.011551505327224751,
|
| 1097 |
+
"aurc": 0.021329015107119983,
|
| 1098 |
"acc_at_80": 0.975,
|
| 1099 |
"acc_at_50": 0.98,
|
| 1100 |
"chance": 0.5,
|
| 1101 |
+
"mean_conf": 0.9395062327384949,
|
| 1102 |
+
"sec": 6.0,
|
| 1103 |
"heldout": false
|
| 1104 |
},
|
| 1105 |
"wic": {
|
| 1106 |
"n": 638,
|
| 1107 |
+
"acc": 0.7163009404388715,
|
| 1108 |
+
"nll": 0.5828382968902588,
|
| 1109 |
+
"brier": 0.38676366209983826,
|
| 1110 |
+
"ece": 0.05568265924259413,
|
| 1111 |
+
"aurc": 0.20020452896287377,
|
| 1112 |
+
"acc_at_80": 0.7647058823529411,
|
| 1113 |
+
"acc_at_50": 0.8119122257053292,
|
| 1114 |
"chance": 0.5,
|
| 1115 |
+
"mean_conf": 0.7616857290267944,
|
| 1116 |
+
"sec": 1.1,
|
| 1117 |
"heldout": false
|
| 1118 |
},
|
| 1119 |
"multirc": {
|
| 1120 |
"n": 1500,
|
| 1121 |
+
"acc": 0.8913333333333333,
|
| 1122 |
+
"nll": 0.300986111164093,
|
| 1123 |
+
"brier": 0.16819369792938232,
|
| 1124 |
+
"ece": 0.03357027542591094,
|
| 1125 |
+
"aurc": 0.03652944216687471,
|
| 1126 |
+
"acc_at_80": 0.9408333333333333,
|
| 1127 |
+
"acc_at_50": 0.9693333333333334,
|
| 1128 |
"chance": 0.5,
|
| 1129 |
+
"mean_conf": 0.9222791194915771,
|
| 1130 |
+
"sec": 6.7,
|
| 1131 |
"heldout": false
|
| 1132 |
},
|
| 1133 |
"cb": {
|
| 1134 |
"n": 56,
|
| 1135 |
"acc": 0.8928571428571429,
|
| 1136 |
+
"nll": 0.3820924460887909,
|
| 1137 |
+
"brier": 0.17802833020687103,
|
| 1138 |
+
"ece": 0.08807407692074777,
|
| 1139 |
+
"aurc": 0.023497414016688704,
|
| 1140 |
+
"acc_at_80": 0.9555555555555556,
|
| 1141 |
"acc_at_50": 0.9642857142857143,
|
| 1142 |
"chance": 0.3333333333333333,
|
| 1143 |
+
"mean_conf": 0.8515821695327759,
|
| 1144 |
"sec": 0.2,
|
| 1145 |
"heldout": true
|
| 1146 |
},
|
| 1147 |
"fever": {
|
| 1148 |
"n": 1500,
|
| 1149 |
+
"acc": 0.8866666666666667,
|
| 1150 |
+
"nll": 0.3317285180091858,
|
| 1151 |
+
"brier": 0.1768377274274826,
|
| 1152 |
+
"ece": 0.017198302884896575,
|
| 1153 |
+
"aurc": 0.03783623559244957,
|
| 1154 |
+
"acc_at_80": 0.9383333333333334,
|
| 1155 |
+
"acc_at_50": 0.972,
|
| 1156 |
"chance": 0.3333333333333333,
|
| 1157 |
+
"mean_conf": 0.8916481137275696,
|
| 1158 |
+
"sec": 3.3,
|
| 1159 |
"heldout": false
|
| 1160 |
},
|
| 1161 |
"wiki_qa": {
|
| 1162 |
"n": 879,
|
| 1163 |
+
"acc": 0.8987485779294653,
|
| 1164 |
+
"nll": 0.2437155544757843,
|
| 1165 |
+
"brier": 0.1463335007429123,
|
| 1166 |
+
"ece": 0.012018481686931658,
|
| 1167 |
+
"aurc": 0.02256072022253832,
|
| 1168 |
+
"acc_at_80": 0.9587482219061166,
|
| 1169 |
+
"acc_at_50": 0.9863636363636363,
|
| 1170 |
"chance": 0.5,
|
| 1171 |
+
"mean_conf": 0.8944457769393921,
|
| 1172 |
+
"sec": 4.3,
|
| 1173 |
"heldout": false
|
| 1174 |
},
|
| 1175 |
"msmarco_rel": {
|
| 1176 |
"n": 1446,
|
| 1177 |
+
"acc": 0.6763485477178424,
|
| 1178 |
+
"nll": 0.6001450419425964,
|
| 1179 |
+
"brier": 0.4146299362182617,
|
| 1180 |
+
"ece": 0.05068599555337411,
|
| 1181 |
+
"aurc": 0.2088099598857701,
|
| 1182 |
+
"acc_at_80": 0.7113223854796888,
|
| 1183 |
"acc_at_50": 0.7745504840940526,
|
| 1184 |
"chance": 0.5,
|
| 1185 |
+
"mean_conf": 0.724265456199646,
|
| 1186 |
+
"sec": 3.4,
|
| 1187 |
"heldout": false
|
| 1188 |
},
|
| 1189 |
"medmcqa": {
|
| 1190 |
"n": 1500,
|
| 1191 |
+
"acc": 0.5006666666666667,
|
| 1192 |
+
"nll": 1.1471084356307983,
|
| 1193 |
+
"brier": 0.6143077611923218,
|
| 1194 |
+
"ece": 0.11058435388406118,
|
| 1195 |
+
"aurc": 0.30678103490481,
|
| 1196 |
+
"acc_at_80": 0.5591666666666667,
|
| 1197 |
+
"acc_at_50": 0.6693333333333333,
|
| 1198 |
"chance": 0.25,
|
| 1199 |
+
"mean_conf": 0.6112509965896606,
|
| 1200 |
+
"sec": 2.7,
|
| 1201 |
"heldout": false
|
| 1202 |
},
|
| 1203 |
"quality": {
|
| 1204 |
"n": 1500,
|
| 1205 |
+
"acc": 0.5146666666666667,
|
| 1206 |
+
"nll": 1.3762184381484985,
|
| 1207 |
+
"brier": 0.6870999932289124,
|
| 1208 |
+
"ece": 0.2046667002439499,
|
| 1209 |
+
"aurc": 0.34077759893205944,
|
| 1210 |
+
"acc_at_80": 0.5633333333333334,
|
| 1211 |
+
"acc_at_50": 0.6173333333333333,
|
| 1212 |
"chance": 0.25,
|
| 1213 |
+
"mean_conf": 0.7156844139099121,
|
| 1214 |
+
"sec": 18.7,
|
| 1215 |
"heldout": true
|
| 1216 |
},
|
| 1217 |
"xstory_cloze": {
|
| 1218 |
"n": 1500,
|
| 1219 |
+
"acc": 0.9613333333333334,
|
| 1220 |
+
"nll": 0.1043626070022583,
|
| 1221 |
+
"brier": 0.05890524387359619,
|
| 1222 |
+
"ece": 0.006435122688611336,
|
| 1223 |
+
"aurc": 0.004630966744998978,
|
| 1224 |
+
"acc_at_80": 0.9966666666666667,
|
| 1225 |
"acc_at_50": 0.9986666666666667,
|
| 1226 |
"chance": 0.5,
|
| 1227 |
+
"mean_conf": 0.9595363140106201,
|
| 1228 |
+
"sec": 3.0,
|
| 1229 |
"heldout": true
|
| 1230 |
},
|
| 1231 |
"abstain_probe": {
|
| 1232 |
"n": 1500,
|
| 1233 |
+
"acc": 0.794,
|
| 1234 |
+
"nll": 0.6064885258674622,
|
| 1235 |
+
"brier": 0.2841270864009857,
|
| 1236 |
+
"ece": 0.037370719770590435,
|
| 1237 |
+
"aurc": 0.054226749633954824,
|
| 1238 |
+
"acc_at_80": 0.8966666666666666,
|
| 1239 |
+
"acc_at_50": 0.968,
|
| 1240 |
"chance": 0.17950317460317464,
|
| 1241 |
+
"mean_conf": 0.8279028534889221,
|
| 1242 |
+
"sec": 3.8,
|
| 1243 |
"heldout": true
|
| 1244 |
},
|
| 1245 |
"toolace": {
|
| 1246 |
"n": 1000,
|
| 1247 |
+
"acc": 0.924,
|
| 1248 |
+
"nll": 0.24141927063465118,
|
| 1249 |
+
"brier": 0.11758895963430405,
|
| 1250 |
+
"ece": 0.025210315197706203,
|
| 1251 |
+
"aurc": 0.02069434380751835,
|
| 1252 |
"acc_at_80": 0.97125,
|
| 1253 |
+
"acc_at_50": 0.982,
|
| 1254 |
"chance": 0.12470555555555554,
|
| 1255 |
+
"mean_conf": 0.9345914125442505,
|
| 1256 |
+
"sec": 9.7,
|
| 1257 |
+
"heldout": false
|
| 1258 |
+
},
|
| 1259 |
+
"agenttraj": {
|
| 1260 |
+
"n": 1500,
|
| 1261 |
+
"acc": 0.916,
|
| 1262 |
+
"nll": 0.20774629712104797,
|
| 1263 |
+
"brier": 0.10922007262706757,
|
| 1264 |
+
"ece": 0.0139970025122166,
|
| 1265 |
+
"aurc": 0.00845390614589135,
|
| 1266 |
+
"acc_at_80": 0.9891666666666666,
|
| 1267 |
+
"acc_at_50": 0.9986666666666667,
|
| 1268 |
+
"chance": 0.2,
|
| 1269 |
+
"mean_conf": 0.9180088043212891,
|
| 1270 |
+
"sec": 6.6,
|
| 1271 |
"heldout": false
|
| 1272 |
+
},
|
| 1273 |
+
"mind2web": {
|
| 1274 |
+
"n": 1500,
|
| 1275 |
+
"acc": 0.8086666666666666,
|
| 1276 |
+
"nll": 0.5181217789649963,
|
| 1277 |
+
"brier": 0.2681422531604767,
|
| 1278 |
+
"ece": 0.025145165185133628,
|
| 1279 |
+
"aurc": 0.05308884826749579,
|
| 1280 |
+
"acc_at_80": 0.8883333333333333,
|
| 1281 |
+
"acc_at_50": 0.9666666666666667,
|
| 1282 |
+
"chance": 0.16666666666666666,
|
| 1283 |
+
"mean_conf": 0.8295513391494751,
|
| 1284 |
+
"sec": 4.2,
|
| 1285 |
+
"heldout": false
|
| 1286 |
+
},
|
| 1287 |
+
"synth": {
|
| 1288 |
+
"n": 300,
|
| 1289 |
+
"acc": 0.7566666666666667,
|
| 1290 |
+
"nll": 0.542948305606842,
|
| 1291 |
+
"brier": 0.3258218467235565,
|
| 1292 |
+
"ece": 0.06296401838461561,
|
| 1293 |
+
"aurc": 0.09320432439081776,
|
| 1294 |
+
"acc_at_80": 0.825,
|
| 1295 |
+
"acc_at_50": 0.9266666666666666,
|
| 1296 |
+
"chance": 0.37149999999999994,
|
| 1297 |
+
"mean_conf": 0.7900449633598328,
|
| 1298 |
+
"sec": 0.5,
|
| 1299 |
+
"heldout": false,
|
| 1300 |
+
"per_q": [
|
| 1301 |
+
0.7133333333333334,
|
| 1302 |
+
0.8
|
| 1303 |
+
]
|
| 1304 |
+
},
|
| 1305 |
+
"games": {
|
| 1306 |
+
"n": 500,
|
| 1307 |
+
"acc": 0.998,
|
| 1308 |
+
"nll": 0.0028993848245590925,
|
| 1309 |
+
"brier": 0.0012032075319439173,
|
| 1310 |
+
"ece": 0.002253464698791504,
|
| 1311 |
+
"aurc": 4e-06,
|
| 1312 |
+
"acc_at_80": 1.0,
|
| 1313 |
+
"acc_at_50": 1.0,
|
| 1314 |
+
"chance": 0.2885,
|
| 1315 |
+
"mean_conf": 0.9976871609687805,
|
| 1316 |
+
"sec": 1.2,
|
| 1317 |
+
"heldout": false
|
| 1318 |
+
},
|
| 1319 |
+
"mario": {
|
| 1320 |
+
"n": 276,
|
| 1321 |
+
"acc": 0.9927536231884058,
|
| 1322 |
+
"nll": 0.013057849369943142,
|
| 1323 |
+
"brier": 0.006284656934440136,
|
| 1324 |
+
"ece": 0.010539730821830639,
|
| 1325 |
+
"aurc": 3.9430219014339996e-05,
|
| 1326 |
+
"acc_at_80": 1.0,
|
| 1327 |
+
"acc_at_50": 1.0,
|
| 1328 |
+
"chance": 0.3333333333333333,
|
| 1329 |
+
"mean_conf": 0.9905721545219421,
|
| 1330 |
+
"sec": 6.4,
|
| 1331 |
+
"heldout": false,
|
| 1332 |
+
"per_q": [
|
| 1333 |
+
0.9855072463768116,
|
| 1334 |
+
1.0
|
| 1335 |
+
]
|
| 1336 |
}
|
| 1337 |
},
|
| 1338 |
"agg": {
|
| 1339 |
"in_task": {
|
| 1340 |
+
"acc": 0.8126446732755706,
|
| 1341 |
+
"nll": 0.4534154008986239,
|
| 1342 |
+
"brier": 0.25158945492306806,
|
| 1343 |
+
"ece": 0.03454096135309228,
|
| 1344 |
+
"aurc": 0.09352685127729836,
|
| 1345 |
+
"acc_at_80": 0.8639655870629418,
|
| 1346 |
+
"chance": 0.32498304691713953
|
| 1347 |
},
|
| 1348 |
"heldout": {
|
| 1349 |
+
"acc": 0.7479565063694622,
|
| 1350 |
+
"nll": 0.656638503074646,
|
| 1351 |
+
"brier": 0.35096842001961626,
|
| 1352 |
+
"ece": 0.08591399023846887,
|
| 1353 |
+
"aurc": 0.13724056983642266,
|
| 1354 |
+
"acc_at_80": 0.8022276412618095,
|
| 1355 |
"chance": 0.32053884112032693
|
| 1356 |
},
|
| 1357 |
+
"n_in": 69,
|
| 1358 |
"n_heldout": 23
|
| 1359 |
},
|
| 1360 |
+
"model": "runs/r7_v4/model",
|
| 1361 |
+
"engine": "compile"
|
| 1362 |
}
|
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 3763692048
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:005b182713b64d6a37f37b30740cd36e46225217c0183798fbc3f278bc7d7da3
|
| 3 |
size 3763692048
|
tokenizer_config.json
CHANGED
|
@@ -9,7 +9,7 @@
|
|
| 9 |
"eos_token": "<|endoftext|>",
|
| 10 |
"errors": "replace",
|
| 11 |
"image_token": "<|image_pad|>",
|
| 12 |
-
"is_local":
|
| 13 |
"local_files_only": false,
|
| 14 |
"model_max_length": 262144,
|
| 15 |
"model_specific_special_tokens": {
|
|
|
|
| 9 |
"eos_token": "<|endoftext|>",
|
| 10 |
"errors": "replace",
|
| 11 |
"image_token": "<|image_pad|>",
|
| 12 |
+
"is_local": true,
|
| 13 |
"local_files_only": false,
|
| 14 |
"model_max_length": 262144,
|
| 15 |
"model_specific_special_tokens": {
|