List your company

Code Generation on MBPP+

85.75AccuracyG-Memory
67.70672.390577.07581.7595Aug 4, 2025Sep 25, 2025Nov 16, 2025Jan 7, 2026Feb 28, 2026Apr 21, 2026Jun 12, 2026

Evaluation results

MethodLinks
G-MemoryBackbone=QWEN3-14B, Framework=DyLAN2026.0585.75
G-MemoryBackbone=QWEN3-14B, Framework=AutoGen2026.0585.23
DECENTMEMBackbone=QWEN3-14B, Framework=AutoGen2026.0584.74
ChatDevBackbone=GEMMA4-E4B, Framework=AutoGen2026.0583.04
G-MemoryBackbone=GEMMA4-E4B, Framework=DyLAN2026.0582.45
DECENTMEMBackbone=QWEN3-14B, Framework=DyLAN2026.0582.39
DECENTMEMBackbone=QWEN3-14B, Framework=AgentNet2026.0582.27
Text-SerializationBackbone=Qwen3-14B, Workers=32026.0681.75
DECENTMEMBackbone=GEMMA4-E4B, Framework=DyLAN2026.0581.49
G-MemoryBackbone=GEMMA4-E4B, Framework=AutoGen2026.0581.17
Parallel-SynthesisBackbone=Qwen3-14B, Workers=3, Merging=Weighted averaging (lambda=0.5)2026.0680.42
ChatDevBackbone=QWEN3-14B, Framework=AutoGen2026.0580.36
DECENTMEMBackbone=GEMMA4-E4B, Framework=AutoGen2026.0579.39
VotingBackbone=Qwen3-14B, Workers=32026.0679.37
ChatDevBackbone=QWEN3-14B, Framework=DyLAN2026.0579.17
DiffMASBackbone=DeepSeek-R1-Distill-Qwen-32B2026.0478.5
G-MemoryBackbone=GEMMA4-E4B, Framework=AgentNet2026.0578.39
KVLINKBackbone=Qwen3-14B, Workers=32026.0678.31
DECENTMEMBackbone=QWEN3-8B, Framework=AgentNet2026.0577.93
MetaGPTBackbone=GEMMA4-E4B, Framework=AutoGen2026.0577.69
DiffMASBackbone=Qwen3-14B2026.0477.2
DECENTMEMBackbone=GEMMA4-E4B, Framework=AgentNet2026.0577.17
MetaGPTBackbone=QWEN3-14B, Framework=DyLAN2026.0576.95
ChatDevBackbone=QWEN3-14B, Framework=AgentNet2026.0576.78
MetaGPTBackbone=QWEN3-14B, Framework=AutoGen2026.0576.37
SingleBackbone=Qwen3-14B, Workers=12026.0676.19
DECENTMEMBackbone=QWEN3-8B, Framework=AutoGen2026.0575.97
PlanningModel=Qwen3-8B2026.0275.9
Primitives-based MASModel=Qwen3-8B2026.0275.9
LatentMASModel=DeepSeek-R1-Distill Qwen-32B2026.0275.7
Primitives-based MASModel=DeepSeek-R1-Distill Qwen-32B2026.0275.7
LatentMASBackbone=Qwen3-14B2026.0475.7
LatentMASBackbone=DeepSeek-R1-Distill-Qwen-32B2026.0475.7
No memoryBackbone=GEMMA4-E4B, Framework=AutoGen2026.0575.52
ChatDevBackbone=GEMMA4-E4B, Framework=DyLAN2026.0575.37
DiffMASModel=Qwen3-8B, Communication Category=Trained Latent Communication2026.0474.8
MetaGPTBackbone=GEMMA4-E4B, Framework=DyLAN2026.0574.69
LatentMASModel=Qwen3-8B2026.0274.6
TextMASModel=DeepSeek-R1-Distill Qwen-32B2026.0274.6
VotingModel=DeepSeek-R1-Distill Qwen-32B2026.0274.6
LatentMASModel=Qwen3-8B, Communication Category=Training-free Latent Communication2026.0474.6
TextMASBackbone=DeepSeek-R1-Distill-Qwen-32B2026.0474.6
G-MemoryBackbone=QWEN3-8B, Framework=DyLAN2026.0574.49
MetaGPTBackbone=GEMMA4-E4B, Framework=AgentNet2026.0574.49
VotingModel=Qwen3-8B2026.0274.3
PlanningModel=DeepSeek-R1-Distill Qwen-32B2026.0274.3
No memoryBackbone=QWEN3-14B, Framework=AutoGen2026.0574.21
MicroMixModel=32B2025.0874.1
G-MemoryBackbone=QWEN3-8B, Framework=AutoGen2026.0573.92
SingleModel=DeepSeek-R1-Distill Qwen-32B2026.0273.8
ReviewModel=DeepSeek-R1-Distill Qwen-32B2026.0273.8
SingleBackbone=DeepSeek-R1-Distill-Qwen-32B2026.0473.8
INT8Model=32B2025.0873.5
Disagreement-Guided Strategy Routing (Qwen2.5-Coder-7B-Instruct)Backbone=Qwen2.5-Coder-7B-Instruct, Sampling strategy=Ours2026.0473.5
INT8Model=14B2025.0873
Qwen2.5-Coder-7B-Instruct (BoN)Backbone=Qwen2.5-Coder-7B-Instruct, Sampling strategy=BoN2026.0473
DECENTMEMBackbone=QWEN3-8B, Framework=DyLAN2026.0572.99
TextMASBackbone=Qwen3-14B2026.0472.8
No memoryBackbone=GEMMA4-E4B, Framework=DyLAN2026.0572.49
ChatDevBackbone=GEMMA4-E4B, Framework=AgentNet2026.0572.18
No memoryBackbone=QWEN3-14B, Framework=DyLAN2026.0572.1
ExPODistillation Setting=Multi-Teacher Distillation2026.0272
DiffMASModel=Qwen3-4B, Communication Category=Trained Latent Communication2026.0471.9
G-MemoryBackbone=QWEN3-14B, Framework=AgentNet2026.0571.29
DiffMASBackbone=Mistral3-8B2026.0471.1
ChatDevBackbone=QWEN3-8B, Framework=AutoGen2026.0571.05
G-MemoryBackbone=QWEN3-8B, Framework=AgentNet2026.0570.94
ChatDevBackbone=QWEN3-8B, Framework=DyLAN2026.0570.94
FP16Model=32B2025.0870.9
ExOPDDistillation Setting=Single-Teacher Distillation2026.0270.7
ExOPDDistillation Setting=Multi-Teacher Distillation2026.0270.6
ReviewModel=Qwen3-8B2026.0270.6
Primitives-based MASModel=DeepSeek-R1-Distill Llama-70B2026.0270.6
LatentMASModel=Qwen3-4B, Communication Category=Training-free Latent Communication2026.0470.6
BinaryModel=Dream-Coder-7B (instruct), NFE budget=1/1, FT steps=-2025.1070.4
No memoryBackbone=GEMMA4-E4B, Framework=AgentNet2026.0570.23
TeacherDistillation Setting=Baseline2026.0270.2
ExPODistillation Setting=Single-Teacher Distillation2026.0270.2
MicroMixModel=14B2025.0870.1
Qwen2.5-Coder-7B-InstructBackbone=Qwen2.5-Coder-7B-Instruct, Sampling strategy=Base2026.0470.1
OPDDistillation Setting=Single-Teacher Distillation2026.0269.9
G-MemoryBackbone=GEMMA4-E2B, Framework=DyLAN2026.0569.89
SingleBackbone=Mistral3-8B2026.0469.8
SFTDistillation Setting=Multi-Teacher Distillation2026.0269.6
SMModel=Dream-Coder-7B (instruct), NFE budget=1/1, FT steps=33.5k2025.1069.6
OPDDistillation Setting=Multi-Teacher Distillation2026.0269.5
TextMASModel=Qwen3-8B2026.0269.5
TextMASModel=Qwen3-8B, Communication Category=Text Communication2026.0469.5
Qwen2.5-Coder-7B-Instruct (Majority)Backbone=Qwen2.5-Coder-7B-Instruct, Sampling strategy=Majority2026.0469.5
TextMASModel=Qwen3-4B, Communication Category=Text Communication2026.0469.3
DECENTMEMBackbone=GEMMA4-E2B, Framework=AutoGen2026.0569.28
FP16Model=14B2025.0869.2
Qwen2.5-Coder-7B-Instruct (DV)Backbone=Qwen2.5-Coder-7B-Instruct, Sampling strategy=DV2026.0469.1
MetaGPTBackbone=QWEN3-8B, Framework=AutoGen2026.0568.99
MetaGPTBackbone=QWEN3-14B, Framework=AgentNet2026.0568.81
TextMASModel=DeepSeek-R1-Distill Llama-70B2026.0268.8
VotingModel=DeepSeek-R1-Distill Llama-70B2026.0268.8
No memoryBackbone=QWEN3-14B, Framework=AgentNet2026.0568.72
SingleBackbone=Qwen3-14B2026.0468.5
Qwen2.5-Coder-7B-Instruct (SCoP)Backbone=Qwen2.5-Coder-7B-Instruct, Sampling strategy=SCoP2026.0468.4