List your company
ResearchFormerly WizWand

Temporal Question Answering on TimeQA Hard

77.7EMDeepSeek-V3-AdapTime
-1.3419.1839.760.22Nov 16, 2023Apr 12, 2024Sep 8, 2024Feb 4, 2025Jul 3, 2025Nov 29, 2025Apr 27, 2026
Updated 4mo ago

Evaluation results

MethodLinks
DeepSeek-V3-AdapTimeBackbone=DeepSeek-V3-0324, Strategy=AdapTime2026.0477.7——79.2
DeepSeek-V3-Step-backBackbone=DeepSeek-V3-0324, Strategy=Step-back2026.0476.4——77.9
DeepSeek-V3-Self-refinementBackbone=DeepSeek-V3-0324, Strategy=Self-refinement2026.0476.4——78.4
DeepSeek-V3-CoTBackbone=DeepSeek-V3-0324, Strategy=Chain-of-Thought (CoT)2026.0475.6——77
DeepSeek-V3-ICLBackbone=DeepSeek-V3-0324, Strategy=In-context Learning (ICL)2026.0468.8——71.6
Qwen-3-8B-AdapTimeBackbone=Qwen-3-8B, Strategy=AdapTime2026.0466.5——68.2
TG-LLMStrategy=Temporal Reasoning2026.0463.1——66.4
Qwen-3-8B-CoTBackbone=Qwen-3-8B, Strategy=Chain-of-Thought (CoT)2026.0462.9——64.7
REMEMO-largeBackbone=REMEMO-large, Strategy=Supervised Fine-tuning2026.0460.5——69.3
T5-largeBackbone=T5-large, Strategy=Supervised Fine-tuning2026.0459.5——68.1
REMEMO-baseBackbone=REMEMO-base, Strategy=Supervised Fine-tuning2026.0458.2——67.3
Qwen-3-8B-ICLBackbone=Qwen-3-8B, Strategy=In-context Learning (ICL)2026.0456.9——60.1
T5-baseBackbone=T5-base, Strategy=Supervised Fine-tuning2026.0455.6——64.1
GPT-4Backbone=GPT-4, Strategy=Vanilla2026.0454.6——57.1
T5-L-FiD-PITBackbone=T5-Large, Framework=Fusion-in-Decoder (FiD), Strategy=PIT2023.1152.747.349.861
T5-L-FiDBackbone=T5-Large, Framework=Fusion-in-Decoder (FiD), Implementation=re-implemented2023.1150.545.147.659.8
T5-B-FiD-PITBackbone=T5-Base, Framework=Fusion-in-Decoder (FiD), Strategy=PIT2023.114641.143.354.7
T5-B-FiDBackbone=T5-Base, Framework=Fusion-in-Decoder (FiD), Implementation=re-implemented2023.1144.339.441.753.2
QAaPStrategy=Embedding-based Reasoning2026.0439.6——49.3
T5-B-PITBackbone=T5-Base, Strategy=PIT2023.113934.236.448.4
T5-BBackbone=T5-Base2023.1137.332.934.946.8
LLaMA-3-8B-AdapTimeBackbone=LLaMA-3.1-8B-Instruct, Strategy=AdapTime2026.0433.3——38.6
LLaMA-3-8B-CoTBackbone=LLaMA-3.1-8B-Instruct, Strategy=Chain-of-Thought (CoT)2026.0431.6——33.8
T5-B-FiDBackbone=T5-Base, Framework=Fusion-in-Decoder (FiD), Source=Chen et al. (2021)2023.1110.3——19.7
LLaMA-3-8B-ICLBackbone=LLaMA-3.1-8B-Instruct, Strategy=In-context Learning (ICL)2026.041.7——3.6
DYNA2026.06—63.24——
DynaGen2026.06—60.56——
RAG2026.06—54.12——