Share your thoughts, 1 month free Claude Pro on usSee more

Language Modeling on LAMBADA (Accuracy)

79.4Accuracy

LLaMA-2 70B

Updated 2d ago

Evaluation Results

Method	Links
LLaMA-2 70B 2026.03		79.4
Focus (LLaMA-2 70B) 2026.03		79.4
LLaMA-2 13B 2026.03		76.6
Focus (LLaMA-2 13B) 2026.03		76.6
Gemma 2 9B 2026.03		75.5
Focus (Gemma 2 9B) 2026.03		75.5
Mistral 7B 2026.03		75.3
Focus (Mistral 7B) 2026.03		75.3
OLMo-2 7B 2026.03		73.2
Focus (OLMo-2 7B) 2026.03		73.2
Qwen 2.5 7B 2026.03		70.7
Focus (Qwen 2.5 7B) 2026.03		70.7
Full 2026.04		64.6
Stochastic 2026.04		64.6
SWA 2026.04		64.6
Stochastic 2026.04		64.6
MoBA (k=2) 2026.04		64.6
SWA 2026.04		64.5
MoBA (k=2) 2026.04		64.5
NITP 2026.05		64.49
IntAttention 2025.11		63.61
FP16 2025.11		62.95
Stochastic 2026.04		62.9
Quant-Only 2025.11		62.62
MoHGE-3B 2026.04		62.37
NTP 2026.05		62.32
HMoE-3B 2026.04		62.25
Llama-3.2-1B 2025.07		62.1
MoDSE-3B 2026.04		61.98
OLMo-1B-0724 2025.07		61
IntAttention 2025.11		58.78
Ettin-Dec-1B 2025.07		58.4
Quant-Only 2025.11		58
FP16 2025.11		57.85
NTP 2026.05		55.45
NITP 2026.05		55.45
MoHGE-1B 2026.04		53.75
SmolLM2-360m 2025.07		53.5
MoE-1B 2026.04		53.2
Ettin-Dec-400m 2025.07		52.3
Dense 2026.04		51.87
Pythia-410m 2025.07		51.5
GPT-2 1.5B 2026.03		51.2
Focus (GPT-2 1.5B) 2026.03		51.2
FP16 2025.11		50.73
NITP 2026.05		50.52
NTP 2026.05		50.36
SWA 2026.04		50.2
BF16 2026.04		49.64
LoRA (r=640) 2026.04		48.8
AdaHOP-Lv2 2026.04		48.57
Pre-proj + skip 2026.04		48.4
HALO 2026.04		48.04
HALO 2026.04		47.95
GPT-2 774M 2026.03		47.7
Focus (GPT-2 774M) 2026.03		47.7
BF16 2026.04		47.53
AdaHOP-Lv1 2026.04		47.41
AdaHOP-Lv2 2026.04		47.28
Pre-proj 2026.04		47
Pre-proj + LoRA 2026.04		47
IntAttention 2025.11		46.94
Baseline 2026.04		46.6
BF16 2026.04		46.26
Quant-Only 2025.11		46.24
HALO 2026.04		45.9
AdaHOP-Lv2 2026.04		45.42
Tseng et al. 2026.04		45.12
AdaHOP-Lv1 2026.04		45.02
MoBA (k=2) 2026.04		44.7
MXFP4+Hadamard 2026.04		43.39
Ettin-Dec-150m 2025.07		43.2
SmolLM2-135m 2025.07		42.9
HALO 2026.04		39.03
AdaHOP-Lv2 2026.04		38.55
AdaHOP-Lv1 2026.04		38.21
BF16 2026.04		38.17
MXFP4+Hadamard 2026.04		37.68
Tseng et al. 2026.04		36.81
Ettin-Dec-68m 2025.07		35.2
Stochastic 2026.04		33.2
Pythia-160m 2025.07		32.9
GPT-2 124M 2026.03		32.6
Focus (GPT-2 124M) 2026.03		32.6
Ettin-Dec-32m 2025.07		28.5
MXFP4 2026.04		27.71
DistilGPT 2025.07		25
Ettin-Dec-17m 2025.07		23
Pre-proj + skip 2026.04		18
Gaussian-Gaussian 2026.05		17.89
Pre-proj + LoRA 2026.04		16.2
Gaussian-Base 2026.05		15.54
Pre-proj 2026.04		15.4
SWA 2026.04		15.1
Baseline 2026.04		12.8
LoRA (r=480) 2026.04		12.6
MXFP4 2026.04		11.93
Gaussian 2026.05		10.25
Pythia-14m 2025.07		7.1
MXFP4 2026.04		6.09

Showing 100 of 103 rows