Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

MuSR

Benchmarks

Task NameDataset NameSOTA ResultTrend
Multistep ReasoningMuSR
Accuracy73.33
53
ReasoningMuSR 0-shot
Reasoning Score (0-shot)48.82
46
Multistep Soft ReasoningMuSR
Accuracy69
41
ReasoningMuSR
Accuracy71.89
32
Multistep ReasoningMUSR
Accuracy61.67
31
Multistep Soft ReasoningMUSR
Accuracy (Multi-choice)50.77
27
Math & LogicMUSR
MUSR Performance42.12
24
Multi-step Narrative ReasoningMUSR
Accuracy65.86
22
ReasoningMuSR (test)
Accuracy73.9
17
Multi-step Soft ReasoningMuSR
Accuracy65.3
12
Murder Mystery ReasoningMUSR murder (test)
Accuracy75
12
Question AnsweringMuSR n=756 (test)
Activation Score98.8
11
Multistep Soft ReasoningMuSR n=756 (test)
Activation (Act)100
11
ReasoningMuSR
MuSR Score56.48
10
Multi-hop ReasoningMuSR
Accuracy43.12
10
ReasoningMuSR
MuSR Score37.14
9
Self-doubt detectionMuSR 90-trace
AUROC (Self-doubt)83.66
7
Adding MistakeMuSR
AOC0.731
7
Truncated CoT AnsweringMuSR
AOC33.6
7
Multi-step narrative inferenceMuSR
Exact Match68.76
6
Multistep ReasoningMUSR team (test)
Accuracy69
6
Multistep ReasoningMUSR object (test)
Accuracy68.9
6
Team Allocation ReasoningMUSR team (test)
Accuracy69
6
Object Placement ReasoningMUSR object (test)
Accuracy68.9
6
Multistep ReasoningMUSR-fr
Average Score33.79
6
Showing 25 of 30 rows