Our new X account is live! Follow @wizwand_team for updates
Home
/
Benchmarks
Multi-turn RL navigation on FrozenLake held-out (val)
Loading...
60.7
Success Rate
TSR (Beam Search)
4.228
18.889
33.55
48.211
Feb 12, 2026
Success Rate
Response Length
Turns
Updated 4d ago
Evaluation Results
Method
Method
Links
Success Rate
Response Length
Turns
TSR (Beam Search)
Model scale=Qwen2.5-3B
2026.02
60.7
98
3.1
TSR (Lookahead)
Model scale=Qwen2.5-3B
2026.02
57
135
3.3
TSR (Best-of-N)
Model scale=Qwen2.5-3B
2026.02
55
161
3.5
Instance Filtering
Model scale=Qwen2.5-3B
2026.02
48.7
174
3.6
TSR (Beam Search)
Model scale=Qwen2.5-0.5B
2026.02
30
152
3.5
TSR (Lookahead)
Model scale=Qwen2.5-0.5B
2026.02
27.8
168
3.8
TSR (Best-of-N)
Model scale=Qwen2.5-0.5B
2026.02
25
182
4.2
Instance Filtering
Model scale=Qwen2.5-0.5B
2026.02
19.7
157
4.1
Base Model
Model scale=Qwen2.5-3B
2026.02
12.7
225
3.9
Base Model
Model scale=Qwen2.5-0.5B
2026.02
6.4
276
4.6
Feedback
Search any
task
Search any
task