List your company
ResearchFormerly WizWand

Truthfulness on TruthfulQA (Metric: TruthfulQA)

59.2TruthfulQADPO + MaPPO
36.94442.72248.554.278Mar 17, 2026Mar 25, 2026Apr 3, 2026Apr 12, 2026Apr 20, 2026Apr 29, 2026May 8, 2026
Updated 6mo ago

Evaluation results

MethodLinks
DPO + MaPPOBase Model=Qwen2.5-7B-Instruct2026.0559.2
DPOBase Model=Qwen2.5-7B-Instruct2026.0557.1
ITBase Model=Qwen2.5-7B-Instruct2026.0556.4
I-DPO + MaPPOBase Model=Qwen2.5-7B-Instruct2026.0556.2
I-DPOBase Model=Qwen2.5-7B-Instruct2026.0555.9
O-LoRAModel=Qwen2.52026.0442.9
Moderate-GiModel=Qwen32026.0442.8
Moderate-GiModel=Qwen2.52026.0442.5
O-LoRAModel=Qwen32026.0442.3
CosineModel=1B, Pre-training Scheduler=Cosine, alpha_pre=0.1, alpha_mid=1.0, Pre-training tokens=2T, Mid-training tokens=500B, Training Stage=Supervised fine-tuned (SFT)2026.0342
Moderate-GiModel=LLaMA-3.12026.0441.5
CosineModel=1B, Pre-training Scheduler=Cosine, alpha_pre=0.1, alpha_mid=0.0, Pre-training tokens=2T, Mid-training tokens=500B, Training Stage=Supervised fine-tuned (SFT)2026.0341.1
EWCModel=Qwen32026.0440.5
WSDModel=1B, Pre-training Scheduler=WSD, alpha_pre=1.0, alpha_mid=0.0, Pre-training tokens=2T, Mid-training tokens=500B, Training Stage=Supervised fine-tuned (SFT)2026.0340.4
RandomModel=Qwen32026.0440.1
O-LoRAModel=LLaMA-3.12026.0440
WSDModel=1B, Pre-training Scheduler=WSD, alpha_pre=0.1, alpha_mid=0.0, Pre-training tokens=2T, Mid-training tokens=500B, Training Stage=Supervised fine-tuned (SFT)2026.0339.8
WSDModel=1B, Pre-training Scheduler=WSD, alpha_pre=0.1, alpha_mid=1.0, Pre-training tokens=2T, Mid-training tokens=500B, Training Stage=Supervised fine-tuned (SFT)2026.0339.7
BaselineModel=Qwen32026.0439.7
KLModel=Qwen32026.0439.7
LinearModel=1B, Pre-training Scheduler=Linear, alpha_pre=0.1, alpha_mid=1.0, Pre-training tokens=2T, Mid-training tokens=500B, Training Stage=Supervised fine-tuned (SFT)2026.0339.4
LinearModel=1B, Pre-training Scheduler=Linear, alpha_pre=0.1, alpha_mid=0.0, Pre-training tokens=2T, Mid-training tokens=500B, Training Stage=Supervised fine-tuned (SFT)2026.0339
RandomModel=LLaMA-3.12026.0438.6
KLModel=LLaMA-3.12026.0438.6
RandomModel=Qwen2.52026.0438.4
BaselineModel=Qwen2.52026.0438.2
Grad. ClipModel=Qwen2.52026.0438.2
Warmup-Stable-Only (WSO)Model=1B, Pre-training Scheduler=Warmup-Stable-Only (WSO), alpha_pre=1.0, alpha_mid=1.0, Pre-training tokens=2T, Mid-training tokens=500B, Training Stage=Supervised fine-tuned (SFT)2026.0338.1
EWCModel=Qwen2.52026.0438.1
EWCModel=LLaMA-3.12026.0438
KLModel=Qwen2.52026.0437.8
BaselineModel=LLaMA-3.12026.0437.8