Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Instruction Following Evaluation on HelpSteer2 (val)
Loading...
71.8
Quality
SEE
56.928
60.789
64.65
68.511
Jun 3, 2026
Quality
Calibration Score
Win-rate Quality
Win-rate Calibration
Updated 1mo ago
Evaluation Results
Method
Method
Links
Quality
Calibration Score
Win-rate Quality
Win-rate Calibration
SEE
Judge=Gemini 3.1 Flash...
2026.06
71.8
70.9
-
-
SEE
unique examples=160, b...
2026.06
70.4
73.1
67.1
70
Adapted RLCR
Judge=Gemini 3.1 Flash...
2026.06
69.1
67
-
-
Qwen 3-4B
Judge=Gemini 3.1 Flash...
2026.06
68.2
65.1
-
-
Adapted RLCR
unique examples=~5,000...
2026.06
66.2
67.5
57
61.7
Qwen3-4B
status=base model
2026.06
64.4
63.2
50
50
SEE
Judge=Claude Sonnet 4.6
2026.06
59.9
67
-
-
Adapted RLCR
Judge=Claude Sonnet 4.6
2026.06
58.1
62.1
-
-
Qwen 3-4B
Judge=Claude Sonnet 4.6
2026.06
57.5
58.5
-
-
Feedback
Search any
task
Search any
task