List your company
ResearchFormerly WizWand

Knowledge-Intensive Reasoning on HotpotQA (F1 score)

0.654F1 ScoreLlama3.1-8B + ARPO
0.074720.225110.37550.52589Sep 28, 2025Nov 8, 2025Dec 19, 2025Jan 29, 2026Mar 11, 2026Apr 21, 2026Jun 1, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Llama3.1-8B + ARPOBase Model=Llama3.1-8B, Algorithm=ARPO2025.120.654———
EAPOBackbone=Llama3.1-8B-Instruct2026.060.62446.71.98—
Reinforce++Backbone=Llama3.1-8B-Instruct2026.060.59243.91.97—
Qwen2.5-7B + GRPOBase Model=Qwen2.5-7B, Algorithm=GRPO2025.120.59———
Qwen2.5-7B + ARPOBase Model=Qwen2.5-7B, Algorithm=ARPO2025.120.588———
GRPOBackbone=Llama3.1-8B-Instruct2026.060.58844.32.89—
Qwen2.5-3B + ARPOBase Model=Qwen2.5-3B, Algorithm=ARPO2025.120.585———
Llama3.1-8B + GRPOBase Model=Llama3.1-8B, Algorithm=GRPO2025.120.578———
Qwen2.5-7B + DAPOBase Model=Qwen2.5-7B, Algorithm=DAPO2025.120.577———
Llama3.1-8B + Reinforce ++Base Model=Llama3.1-8B, Algorithm=Reinforce ++2025.120.571———
Llama3.1-8B + DAPOBase Model=Llama3.1-8B, Algorithm=DAPO2025.120.566———
Qwen2.5-3B + GRPOBase Model=Qwen2.5-3B, Algorithm=GRPO2025.120.565———
Qwen2.5-3B + Reinforce ++Base Model=Qwen2.5-3B, Algorithm=Reinforce ++2025.120.559———
Qwen2.5-7B + Reinforce ++Base Model=Qwen2.5-7B, Algorithm=Reinforce ++2025.120.551———
Qwen2.5-3B + DAPOBase Model=Qwen2.5-3B, Algorithm=DAPO2025.120.548———
GRPOBackbone Model=Qwen2.5-7B-Instruct2026.060.54442.52.62—
EAPOBackbone Model=Qwen2.5-7B-Instruct2026.060.54142.51.99—
Reinforce++Backbone Model=Qwen2.5-7B-Instruct2026.060.506401.81—
Llama3.1-8B + TIR PromptingBase Model=Llama3.1-8B, Algorithm=TIR Prompting2025.120.485———
AEPOBackbone Model=Qwen2.5-7B-Instruct2026.060.477364.26—
EAPOBackbone Model=Qwen2.5-3B-Instruct2026.060.46538.51.84—
ARPOBackbone Model=Qwen2.5-7B-Instruct2026.060.45734.51.93—
ToolStarBackbone Model=Qwen2.5-7B-Instruct2026.060.42432.53.82—
Reinforce++Backbone Model=Qwen2.5-3B-Instruct2026.060.42337.52.99—
ARPOBackbone Model=Qwen2.5-3B-Instruct2026.060.41232.52.28—
GRPOBackbone Model=Qwen2.5-3B-Instruct2026.060.40731.82.7—
ToolStarBackbone Model=Qwen2.5-3B-Instruct2026.060.385302.29—
TIRBackbone=Llama3.1-8B-Instruct2026.060.30729.83.12—
Llama3.1-8BBase Model=Llama3.1-8B, Algorithm=Direct Reasoning2025.120.243———
EAPO2025.090.2375———
BaseBackbone=Llama3.1-8B-Instruct2026.060.2237.2——
Self-Exploratory RL2025.090.2146———
Qwen2.5-3B + TIR PromptingBase Model=Qwen2.5-3B, Algorithm=TIR Prompting2025.120.154———
TIRBackbone Model=Qwen2.5-3B-Instruct2026.060.15410.41.44—
Qwen2.5-7B + TIR PromptingBase Model=Qwen2.5-7B, Algorithm=TIR Prompting2025.120.148———
TIRBackbone Model=Qwen2.5-7B-Instruct2026.060.1386.72.37—
Base Model2025.090.1286———
Qwen2.5-7BBase Model=Qwen2.5-7B, Algorithm=Direct Reasoning2025.120.122———
BaseBackbone Model=Qwen2.5-7B-Instruct2026.060.1065.6——
Qwen2.5-3BBase Model=Qwen2.5-3B, Algorithm=Direct Reasoning2025.120.097———
BaseBackbone Model=Qwen2.5-3B-Instruct2026.060.0977.2——
APPOBackbone=Llama3.1-8B-Instruct2026.06———65.4
APPOBackbone=Qwen2.5-7B-Instruct2026.06———66.5
ARPOBackbone=Llama3.1-8B-Instruct2026.06———65.4
ARPOBackbone=Qwen2.5-7B-Instruct2026.06———58.8
CISPOBackbone=Llama3.1-8B-Instruct2026.06———57.3
CISPOBackbone=Qwen2.5-7B-Instruct2026.06———59.3
DAPOBackbone=Llama3.1-8B-Instruct2026.06———56.6
DAPOBackbone=Qwen2.5-7B-Instruct2026.06———57.7
GIGPOBackbone=Llama3.1-8B-Instruct2026.06———61.8
GIGPOBackbone=Qwen2.5-7B-Instruct2026.06———58.1
GPPOBackbone=Llama3.1-8B-Instruct2026.06———61.8
GPPOBackbone=Qwen2.5-7B-Instruct2026.06———60.7
GRPOBackbone=Llama3.1-8B-Instruct2026.06———57.8
GRPOBackbone=Qwen2.5-7B-Instruct2026.06———59
Reinforce++Backbone=Llama3.1-8B-Instruct2026.06———57.1
Reinforce++Backbone=Qwen2.5-7B-Instruct2026.06———55.1
TIR PromptingBackbone=Llama3.1-8B-Instruct2026.06———48.5
TIR PromptingBackbone=Qwen2.5-7B-Instruct2026.06———14.8
Zero-shotBackbone=Llama3.1-8B-Instruct2026.06———24.3
Zero-shotBackbone=Qwen2.5-7B-Instruct2026.06———12.2