List your company

ResearchFormerly WizWand

Qwen

Benchmarks

TaskDatasetSOTA resultTrendResults

Certifiable Acceptance Length Estimation

Qwen3-4B

100Mean Certifiable Acceptance Length

72

Efficiency Benchmarking

Qwen3-8B Single-layer forward+backward setup

36.6Time (ms)

57

Qwen-1.5B-Instruct 2.5 (test)

93Score S(γ)

56

Model Discovery

Qwen-3B model tree Extended Discovery

233.8Rank

48

Jailbreak Defense

Qwen2-VL

0ASR

36

Toxicity Defense

Qwen2-VL

0.05Toxicity Score

36

Qwen3-8B 2k prompts Decode-heavy workload

46,782Throughput (tok/s)

30

Inference Throughput

Qwen3 Query Projection Module NVIDIA A40

80.63Throughput (k tokens/sec)

30

Attention Operator Throughput

Qwen2.5 72B (64 Q-heads/8 KV-heads/128 Head-dimension)

222.5Attention Throughput (TFLOPS)

29

Qwen3-8B 2k prompts Balanced workload

61,106Throughput (tok/s)

28

Training Throughput Analysis

Qwen 7B 2.5

1,847Training Throughput (tokens/s)

28

Qwen3-8B 2k prompts Prefill-heavy workload

106,324Throughput (tok/s)

26

Multimodal Language Understanding

Qwen3 1.7B

58.65Average Performance

24

Function Module Discovery

Qwen 7B-Instruct 2.5

64.6L(F)

24

Function Module Discovery

Qwen 3B Instruct 2.5

56.9L(F)

24

Function Module Discovery

Qwen2.5-1.5B-Instruct

31.4L(F)

24

Attacker Detection

Qwen3-1.7B target τpool=1.509 (test)

53.9Tau Multiplier (×τ)

22

Model Retrieval

Qwen-7B model tree (test)

1Rank

21

Model Retrieval

Qwen-3B model tree (test)

1Rank

21

Jailbreak Attack

Qwen2.5-7B

0.02Normalized Rate (NR)

20

Jailbreak Attack Transferability

Qwen Target Transferability set 27B

38ASR

19

Language Modeling

QWEN3-like (train)

2.398Loss

19

Interpreting Sparse Key-Value Features

Qwen 1.7B 3

21.69Input Success Rate (%)

18

Persona Discovery

Qwen3-80B Large Target

99Similarity Score

18

Persona Discovery

Qwen3-30B Large Target

99Similarity Score

18