blancsw's picture
|
download
raw
1.54 kB

MIXTURE REPORT — compile_0001 → 0001_sota

N = 57937 · taxonomy sota-taxonomy-v1

SOTA floors

Capability Count % Floor Status
reasoning-bearing 48986 84.55% ≥30% pass_overindexed
tool-use 5952 10.27% ≥13% fail_below_floor
long-context 1275 2.20% ≥4% fail_below_floor
science 11307 19.52% ≥4% pass
multilingual 436 0.75% ≥5% fail_below_floor
single-turn 50221 86.68% ≤65% fail_above_ceiling

HARD_GAPS: tools, long_context, multilingual, single_turn

Normalized domains

Domain Count %
math 15072 26.01%
code 14719 25.41%
reasoning 10463 18.06%
instruction 8524 14.71%
agent_tools 5952 10.27%
long_context 1263 2.18%
science 844 1.46%
dialogue 664 1.15%
multilingual 436 0.75%

Teachers

Teacher Count %
qwen3.8-max-preview 48283 83.34%
glm-5.2 5307 9.16%
kimi-code/k3 4347 7.50%

Capabilities

Tag Count %
single_turn 50221 86.68%
reasoning 48986 84.55%
math 15072 26.01%
code 14719 25.41%
science 11307 19.52%
instruction 8524 14.71%
multi_turn 7716 13.32%
tools 5952 10.27%
agent_multi_turn 5909 10.20%
long_context_provisional 1275 2.20%
long_context 1263 2.18%
dialogue 664 1.15%
multilingual 436 0.75%

Xet Storage Details

Size:
1.54 kB
·
Xet hash:
64f0ce53d3bf95ceaee279f3cd945c6a7fe68a158106d643d0211f9b2c8e1cb4

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.