List your company

Question Answering on NQ, TriviaQA, and WebQ (test)

65.5AccuracyChatQA + DTA
41.26847.55953.8560.141May 27, 2025

Evaluation results

MethodLinks
ChatQA + DTABackbone=Llama-3-8b, Abstention Strategy=Divide-Then-Align2025.0565.564.567.265.862.848.967.961.864.7
Ret-Robust + DTABackbone=Llama-2-13b, Abstention Strategy=Divide-Then-Align2025.0564.867.965.366.676.845.556.763.559.9
RAAT + DTABackbone=Llama-2-7b, Abstention Strategy=Divide-Then-Align2025.0564.163.765.564.668.952.86561.763.3
Ret-Robust + ConsistencyBackbone=Llama-2-13b, Abstention Strategy=Consistency2025.0553.971.85461.789.646.46.352.511.2
Ret-Robust + LogitsBackbone=Llama-2-13b, Abstention Strategy=Logits2025.0553.67053.660.787.943.41052.916.9
Ret-Robust + ICLBackbone=Llama-2-13b, Abstention Strategy=ICL2025.055271.65260.389.146.6000
Ret-RobustBackbone=Llama-2-13b2025.0551.67151.659.89044.5000
RAAT + ConsistencyBackbone=Llama-2-7b, Abstention Strategy=Consistency2025.0551.46950.758.582.158.816.358.425.4
Ret-Robust + P(true)Backbone=Llama-2-13b, Abstention Strategy=P(true)2025.0550.95658.557.274.829.737.533.635.4
ChatQA + P(true)Backbone=Llama-3-8b, Abstention Strategy=P(true)2025.0550.145.255.649.946.229.161.942.650.5
RAAT + LogitsBackbone=Llama-2-7b, Abstention Strategy=Logits2025.0549.258.850.554.369.84730.945.136.6
Llama-2-13b OriginalBackbone=Llama-2-13b2025.0548.166.348.155.882.140.7000
RAAT + ICLBackbone=Llama-2-7b, Abstention Strategy=ICL2025.0546.871.246.856.584.460.2000
ChatQA + LogitsBackbone=Llama-3-8b, Abstention Strategy=Logits2025.0546.657.846.851.75144.819.344.927
ChatQA + ConsistencyBackbone=Llama-3-8b, Abstention Strategy=Consistency2025.0546.56146.752.958.746.611.34418
RAATBackbone=Llama-2-7b2025.0546.270.246.255.776.361.7000
ChatQABackbone=Llama-3-8b2025.0546.160.94551.854.546.810.271.817.8
RAAT + P(true)Backbone=Llama-2-7b, Abstention Strategy=P(true)2025.0545654653.868.957.46.7132.111
Llama-3-8b OriginalBackbone=Llama-3-8b2025.0543.96243.951.47642000
ChatQA + ICLBackbone=Llama-3-8b, Abstention Strategy=ICL2025.0543.35541.447.250.340.715.175.425.1
Llama-2-7b OriginalBackbone=Llama-2-7b2025.0542.264.142.250.985.849.9000