List your company

Language Understanding on MMLU (Accuracy, Throughput, and Latency)

84.9AccuracyVerify-Only
43.71654.40865.175.792Dec 2, 2025Dec 3, 2025Dec 5, 2025Dec 7, 2025Dec 8, 2025Dec 10, 2025Dec 12, 2025

Evaluation results

MethodLinks
Verify-OnlyDraft-Target Model Pair=Qwen 2.5 72B – Qwen 2.5 7B2025.1284.910.9520.4081.042
TargetDraft-Target Model Pair=Qwen 2.5 72B – Qwen 2.5 7B2025.1284.54.69—0.446
Gen-OnlyDraft-Target Model Pair=Qwen 2.5 72B – Qwen 2.5 7B2025.1284.110.735—1.021
VanillaDraft-Target Model Pair=Llama 3.1 70B – Llama 3.2 1B2025.1283.88.746—1
VanillaDraft-Target Model Pair=Qwen 2.5 72B – Qwen 2.5 7B2025.1283.710.511—1
AssistedGenDraft-Target Model Pair=Qwen 2.5 72B – Qwen 2.5 7B2025.1283.711.436—1.088
TargetDraft-Target Model Pair=Llama 3.1 70B – Llama 3.1 8B2025.1283.64.376—0.553
TargetDraft-Target Model Pair=Llama 3.1 70B – Llama 3.2 1B2025.1283.64.376—0.5
VanillaDraft-Target Model Pair=Llama 3.1 70B – Llama 3.1 8B2025.1283.57.912—1
AssistedGenDraft-Target Model Pair=Llama 3.1 70B – Llama 3.1 8B2025.1283.57.97—1.007
Gen-OnlyDraft-Target Model Pair=Llama 3.1 70B – Llama 3.2 1B2025.1283.59.085—1.039
AdaSDDraft-Target Model Pair=Qwen 2.5 72B – Qwen 2.5 7B2025.1283.511.3950.4071.084
Verify-OnlyDraft-Target Model Pair=Llama 3.1 70B – Llama 3.1 8B2025.1283.38.4810.4181.072
AssistedGenDraft-Target Model Pair=Llama 3.1 70B – Llama 3.2 1B2025.1283.38.92—1.02
Gen-OnlyDraft-Target Model Pair=Llama 3.1 70B – Llama 3.1 8B2025.1283.18.023—1.014
AdaSDDraft-Target Model Pair=Llama 3.1 70B – Llama 3.1 8B2025.1282.28.6230.4191.09
Verify-OnlyDraft-Target Model Pair=Llama 3.1 70B – Llama 3.2 1B2025.1281.69.3850.4651.073
AdaSDDraft-Target Model Pair=Llama 3.1 70B – Llama 3.2 1B2025.1281.39.9370.4631.136
Cosine (7.5, 0)Schedule Type=Cosine, Thresholds (τ_high, τ_low)=(7.5, 0)2025.1280.81——1.56
Linear (7.5, 0)Schedule Type=Linear, Thresholds (τ_high, τ_low)=(7.5, 0)2025.1280.81——1.56
Exp-k=2 (7.5, 0)Schedule Type=Exponential, k=2, Thresholds (τ_high, τ_low)=(7.5, 0)2025.1280.77——1.62
Linear (7.5, 2.5)Schedule Type=Linear, Thresholds (τ_high, τ_low)=(7.5, 2.5)2025.1280.76——1.43
Cosine (7.5, 2.5)Schedule Type=Cosine, Thresholds (τ_high, τ_low)=(7.5, 2.5)2025.1280.75——1.43
BaselineSchedule=Standard2025.1280.69——1
Exp-k=2 (7.5, 2.5)Schedule Type=Exponential, k=2, Thresholds (τ_high, τ_low)=(7.5, 2.5)2025.1280.69——1.54
Exp-k=16 (7.5, 0)Schedule Type=Exponential, k=16, Thresholds (τ_high, τ_low)=(7.5, 0)2025.1279.31——5
Exp-k=16 (7.5, 2.5)Schedule Type=Exponential, k=16, Thresholds (τ_high, τ_low)=(7.5, 2.5)2025.1279.31——2.5
ProphetSchedule=Prophet2025.1278.73——1.14
DraftDraft-Target Model Pair=Qwen 2.5 72B – Qwen 2.5 7B2025.1272.538.544—3.667
DraftDraft-Target Model Pair=Llama 3.1 70B – Llama 3.1 8B2025.1269.533.274—4.206
DraftDraft-Target Model Pair=Llama 3.1 70B – Llama 3.2 1B2025.1245.395.141—10.878