List your company

Question Answering on WikiHop (test)

74AccuracyFine-Tuning
-1.418.17537.7557.325Sep 17, 2018Dec 5, 2019Feb 22, 2021May 13, 2022Jul 31, 2023Oct 18, 2024Jan 6, 2026

Evaluation results

MethodLinks
Fine-Tuning% Tuned=100, Backbone=Longformer-base2023.0574——
Coref-GRU2018.0959.3——
MHPGM+NOICNOIC=true2018.0957.9——
MHPGMNOIC=false2018.0957.5——
LATENTGRAPHMEMBackbone=Qwen3-8B2026.0148.81—40.37
PREMemBackbone=Qwen3-8B2026.0145.87—31.55
BiDAF2018.0942.9——
Prefix-Propagation% Tuned=0.05, Backbone=Longformer-base2023.0542.2——
MemGenBackbone=SmolLM3-3B2026.0141.35—42.59
Reasoner-onlyBackbone=Qwen3-8B2026.0140.62—31.55
MemGenBackbone=Qwen3-8B2026.0140.5—40.3
LATENTGRAPHMEMBackbone=SmolLM3-3B2026.0140.03—35.71
Prefix-Tuning% Tuned=0.1, Backbone=Longformer-base2023.0538.9——
PREMemBackbone=SmolLM3-3B2026.0138.5—25.77
LATENTGRAPHMEMBackbone=Qwen2.5-1.5B-Instruct2026.0136.13—25.03
Reasoner-onlyBackbone=Qwen2.5-1.5B-Instruct2026.0134.63—23.57
Reasoner-onlyBackbone=SmolLM3-3B2026.0134.5—24.61
MemGenBackbone=Qwen2.5-1.5B-Instruct2026.0132.37—32.54
THEANINEBackbone=SmolLM3-3B2026.0119.75—13.36
A-MemBackbone=SmolLM3-3B2026.0117.13—28.45
THEANINEBackbone=Qwen3-8B2026.0116.15—10.48
A-MemBackbone=Qwen2.5-1.5B-Instruct2026.0112.12—20.53
ROBERTa PT% Tuned=0.1, Backbone=ROBERTa-base2023.0511.7——
Mem0Backbone=Qwen2.5-1.5B-Instruct2026.0111.63—10.15
THEANINEBackbone=Qwen2.5-1.5B-Instruct2026.0110.25—6.64
PREMemBackbone=Qwen2.5-1.5B-Instruct2026.0110.13—10.78
A-MemBackbone=Qwen3-8B2026.019.38—17.89
RAGBackbone=Qwen3-8B2026.015.12—20.04
Mem0Backbone=Qwen3-8B2026.014.13—3.46
Mem0Backbone=SmolLM3-3B2026.012.5—2.26
RAGBackbone=SmolLM3-3B2026.012.25—16.52
RAGBackbone=Qwen2.5-1.5B-Instruct2026.011.5—8.05
BIGBIRD-ETCSetup=Fine-tuning2020.07—82.3—
LongformerSetup=Fine-tuning2020.07—81.9—
MRC-GCNSetup=Fine-tuning2020.07—78.3—
MultiHopSetup=Fine-tuning2020.07—76.5—