List your company
ResearchFormerly WizWand

Truthful Question Answering on TruthfulQA (Accuracy)

56.8Accuracy (TruthfulQA)LLaDA
18.00828.07938.1548.221Apr 17, 2025Jun 24, 2025Sep 1, 2025Nov 9, 2025Jan 16, 2026Mar 26, 2026Jun 3, 2026
Updated 4mo ago

Evaluation results

MethodLinks
LLaDABudget=1/1, Evaluation Protocol=fully generative2026.0556.8
ME-DLM Stage 3Budget=1/2, Evaluation Protocol=fully generative2026.0554.6
IPObackbone=DS-8B2025.0953.9
ME-DLM Stage 3Budget=1/1, Evaluation Protocol=fully generative2026.0553.5
RealSafebackbone=DS-8B, alignment=SFT-based2025.0953.4
ME-DLM Stage 2Budget=1/1, Evaluation Protocol=fully generative2026.0552.5
LLaDABudget=1/2, Evaluation Protocol=fully generative2026.0549.8
ME-DLM Stage 3Budget=1/4, Evaluation Protocol=fully generative2026.0549.8
FPModel=DREAM, Precision=Full Precision2026.0649.76
STaR-QuantModel=LLADA-1.5, Quantization=W8A82026.0649.72
STaR-QuantModel=DREAM, Quantization=W8A82026.0649.02
ME-DLM Stage 2Budget=1/2, Evaluation Protocol=fully generative2026.0548.6
STaR-QuantModel=LLADA, Quantization=W8A82026.0648.56
FPModel=LLADA, Precision=Full Precision2026.0647.49
FPModel=LLADA-1.5, Precision=Full Precision2026.0647.2
ME-DLM Stage 3Budget=1/8, Evaluation Protocol=fully generative2026.0546.9
Basebackbone=DS-8B2025.0946.7
LLaDABudget=1/4, Evaluation Protocol=fully generative2026.0544.4
Qwen2.5Size=490M2025.0439.74
CLIMBSize=950M2025.0439.06
SmolLMSize=360M2025.0437.93
Llama-3.2Size=1.2B2025.0437.67
TinyLlamaSize=1.1B2025.0437.6
CLIMBSize=350M2025.0436.86
ME-DLM Stage 2Budget=1/4, Evaluation Protocol=fully generative2026.0536.1
AMD-OLMoSize=1.2B2025.0432.22
DATEDGPT-INSTRUCT-20240-shot=true, Training data cutoff year=20242026.0328.6
SmolLM-1.7B-Instruct0-shot=true2026.0327.7
DATEDGPT-INSTRUCT-20170-shot=true, Training data cutoff year=20172026.0326.8
DATEDGPT-INSTRUCT-20210-shot=true, Training data cutoff year=20212026.0326.4
TinyLlama-1.1B0-shot=true2026.0326
DATEDGPT-INSTRUCT-20130-shot=true, Training data cutoff year=20132026.0325.1
OPT-1.3B0-shot=true2026.0323.8
Pythia-1B0-shot=true2026.0323.6
LLaDABudget=1/8, Evaluation Protocol=fully generative2026.0523.3
GPT2-XL0-shot=true2026.0322.4
ME-DLM Stage 2Budget=1/8, Evaluation Protocol=fully generative2026.0519.5