List your company
ResearchFormerly WizWand

Reading Comprehension on C3

93.5AccuracyInternLM2-Chat-20B-SFT
25.598443.226760.85578.4833Mar 26, 2024Aug 4, 2024Dec 14, 2024Apr 24, 2025Sep 3, 2025Jan 12, 2026May 24, 2026
Updated 5mo ago

Evaluation results

MethodLinks
InternLM2-Chat-20B-SFTParameter Scale=13~20B, Evaluation Setup=5-shot, Training Protocol=SFT2024.0393.5
InternLM2-Chat-20BParameter Scale=13~20B, Evaluation Setup=5-shot, Training Protocol=RLHF2024.0393.5
InternLM2-Chat-7B-SFTParameter Scale=< 7B, Evaluation Setup=5-shot, Training Protocol=SFT2024.0391.7
InternLM2-Chat-7BParameter Scale=< 7B, Evaluation Setup=5-shot, Training Protocol=RLHF2024.0391.5
Qwen-14B-ChatParameter Scale=13~20B, Evaluation Setup=5-shot2024.0391.5
Qwen2.52025.1284.7
Qwen-7B-ChatParameter Scale=< 7B, Evaluation Setup=5-shot2024.0384.4
Baichuan2-13B-ChatParameter Scale=13~20B, Evaluation Setup=5-shot2024.0384.4
Mixtral-8x7B-Instruct-v0.1Parameter Scale=13~20B, Evaluation Setup=5-shot2024.0382.1
Qwen32025.1281
Qwen-14BModel size=13 ~ 20B, shot=5-shot2024.0380.6
ChatGLM3-6BParameter Scale=< 7B, Evaluation Setup=5-shot2024.0379.3
InternLM2-20BModel size=13 ~ 20B, shot=5-shot2024.0379.1
ChatGLM3-6B-BaseModel size=< 7B, shot=5-shot2024.0378.9
Baichuan2-7B-ChatParameter Scale=< 7B, Evaluation Setup=5-shot2024.0378.5
Qwen-1.5 14BModel Type=Teacher2024.0777.38
Qwen-1.5 7BRole=Teacher2024.0776.03
Gamayun2025.1275.29
InternLM2-7BModel size=< 7B, shot=5-shot2024.0374.1
Qwen-7BModel size=< 7B, shot=5-shot2024.0371.4
KDTeacher Model=Qwen-1.5 14B, Student Model=Qwen-1.5 4B2024.0770.31
DDKTeacher Model=Qwen-1.5 14B, Student Model=Qwen-1.5 4B2024.0770.25
MiniLLMTeacher Model=Qwen-1.5 14B, Student Model=Qwen-1.5 4B2024.0768.78
Llama3.22025.1267.85
CPTTeacher Model=Qwen-1.5 14B, Student Model=Qwen-1.5 4B2024.0767.72
InternLM2-20B-BaseModel size=13 ~ 20B, shot=5-shot2024.0367.6
Mistral-7B-Instruct-v0.2Parameter Scale=< 7B, Evaluation Setup=5-shot2024.0366.9
Baichuan2-13B-BaseModel size=13 ~ 20B, shot=5-shot2024.0365.6
Qwen-1.5 4BModel Type=Student2024.0765.26
Baichuan2-7B-BaseModel size=< 7B, shot=5-shot2024.0364.6
NITPModel Scale=3B2026.0563.67
Engram-27BShots=0-shot2026.0163.6
DDKBase Model=Qwen-1.5 1.8B, Teacher=Qwen-1.5 7B2024.0763.37
NITPModel scale=9bA1b, Evaluation=few-shot, Context length=81922026.0563.01
InternLM2-7B-BaseModel size=< 7B, shot=5-shot2024.0361.9
Engram-40BShots=0-shot2026.0161.8
MiniLLMBase Model=Qwen-1.5 1.8B, Teacher=Qwen-1.5 7B2024.0761.45
KDBase Model=Qwen-1.5 1.8B, Teacher=Qwen-1.5 7B2024.0761.18
CPTBase Model=Qwen-1.5 1.8B, Teacher=Qwen-1.5 7B2024.0760.3
MoE-27BShots=0-shot2026.0160.1
Mixtral-8x7B-v0.1Model size=13 ~ 20B, shot=5-shot2024.0359.1
NTPModel Scale=3B2026.0559.01
Qwen-1.5 1.8BRole=Student2024.0758.27
Dense-4BShots=0-shot2026.0157.7
Gemma32025.1257.31
Llama2-13B-ChatParameter Scale=13~20B, Evaluation Setup=5-shot2024.0356.9
NTPModel scale=9bA1b, Evaluation=few-shot, Context length=81922026.0556.65
Mistral-7B-v0.1Model size=< 7B, shot=5-shot2024.0354.6
NITPModel Scale=2B2026.0553.42
POEMParadigm=TTA2026.0453.41
GPT-3.5Parameter Scale=API, Evaluation Setup=5-shot2024.0352.5
Llama2-7B-ChatParameter Scale=< 7B, Evaluation Setup=5-shot2024.0351.7
T^2ARD+MPTParadigm=MOA2026.0451.2
ATTEMPTParadigm=MTL2026.0451.12
SEEDBackbone=LLaMA2-7B, Selection Ratio=5%2026.0551.1
SyCoParadigm=MOA2026.0451
FOAParadigm=TTA2026.0450.87
MPTParadigm=MTL2026.0450.64
POEM+MPTParadigm=MOA2026.0450.52
SPoTParadigm=MTL2026.0450.3
ClusT3+ATTEMPTParadigm=MOA2026.0450.18
ClusT3Paradigm=TTA2026.0449.32
PTParadigm=Vanilla2026.0449.27
NTPModel Scale=2B2026.0549.26
TENT+SPoTParadigm=MOA2026.0449.25
T^2ARDParadigm=TTA2026.0449
TENTParadigm=TTA2026.0448.18
FULLBackbone=LLaMA2-7B, Selection Ratio=100%2026.0548.1
Llama2-13BModel size=13 ~ 20B, shot=5-shot2024.0347.5
T^2ARD+TA-LoRAParadigm=MOA2026.0447.38
POEM+TA-LoRAParadigm=MOA2026.0446.58
RANDOMBackbone=LLaMA2-7B, Selection Ratio=5%2026.0545.3
NITPModel scale=3bA0.5b, Evaluation=few-shot, Context length=81922026.0544.38
Llama2-7BModel size=< 7B, shot=5-shot2024.0343.8
Llama 2 7B (baseline)Base Model=Llama 2 7B, Compression Ratio=0%2025.0543.8
LLM-StreamlineTrain-Free=false, Base Model=Llama 2 7B, Compression Ratio=25%2025.0543.3
ReplaceMe (Cosine)Train-Free=true, Base Model=Llama 2 7B, Compression Ratio=25%2025.0542.5
TA-LoRAParadigm=MTL2026.0442.27
BASEBackbone=LLaMA2-7B, Selection Ratio=0%2026.0542.1
UIDLTrain-Free=false, Base Model=Llama 2 7B, Compression Ratio=25%2025.0540.2
LaCoTrain-Free=false, Base Model=Llama 2 7B, Compression Ratio=25%2025.0539.7
ReplaceMe (LS)Train-Free=true, Base Model=Llama 2 7B, Compression Ratio=25%2025.0539.4
NTPModel scale=3bA0.5b, Evaluation=few-shot, Context length=81922026.0539.06
NTPModel scale=1.9bA0.3b, Evaluation=few-shot, Context length=81922026.0532.21
NITPModel Scale=0.5B2026.0532.1
SliceGPTTrain-Free=false, Base Model=Llama 2 7B, Compression Ratio=25%2025.0531.5
LLMPrunerTrain-Free=false, Base Model=Llama 2 7B, Compression Ratio=25%2025.0529.7
NITPModel scale=1.9bA0.3b, Evaluation=few-shot, Context length=81922026.0529.69
NTPModel Scale=0.5B2026.0528.21