List your company

Question Answering on PopQA (EM, F1, LLM)

46.1EMIGPO (gold-sup.)
-1.386410.941823.2735.5982Jan 25, 2026Feb 10, 2026Feb 27, 2026Mar 16, 2026Apr 2, 2026Apr 19, 2026May 6, 2026

Evaluation results

MethodLinks
IGPO (gold-sup.)Backbone=Qwen3-8B2026.0546.149.8—
IGPO (gold-sup.)Backbone=Qwen3-4B2026.0544.948.6—
SIOPBackbone=Qwen3-8B2026.0544.547.9—
GRPO (gold-sup.)Backbone=Qwen3-4B2026.0543.647.8—
GRPO (gold-sup.)Backbone=Qwen3-8B2026.0541.847.4—
TTRLBackbone=Qwen3-4B2026.0541.645.8—
SIOPBackbone=Qwen3-4B2026.0541.646.5—
TTRLBackbone=Qwen3-8B2026.0541.647.6—
EMPOBackbone=Qwen3-8B2026.0541.545—
EMPOBackbone=Qwen3-4B2026.0541.246—
LLM (with RAG*)Backbone=Mistral-7B-Instruct, Top-k=5, Comp. Rate=-2026.0134.240.1741.05
SeleComBackbone=Mistral-7B-Instruct, Top-k=5, Comp. Rate=82x2026.0133.3240.5139.56
LLM (with RAG*)Backbone=Qwen2.5-7B-Instruct, Top-k=5, Comp. Rate=-2026.0132.2838.5237.92
SeleComBackbone=Qwen2.5-7B-Instruct, Top-k=5, Comp. Rate=82x2026.0132.2338.7139.02
SeleComBackbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=82x2026.0131.4237.6137.65
SeleComBackbone=Qwen2.5-7B-Instruct, Top-k=1, Comp. Rate=82x2026.0131.237.7137.04
LLM (with RAG*)Backbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=-2026.0130.0136.2735.83
LLM (with RAG*)Backbone=Qwen2.5-7B-Instruct, Top-k=1, Comp. Rate=-2026.0129.3335.7735.13
ICAEBackbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=4x2026.0126.8433.836.71
LLM (with RAG)Backbone=Qwen2.5-7B-Instruct, Top-k=1, Comp. Rate=-2026.0125.2233.1938.09
COCOMBackbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=128x2026.0121.7226.2125.32
LLM (with RAG)Backbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=-2026.0120.9129.2638.9
LLMLingua-2Backbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=4x2026.0114.0221.2929.68
LLM (w/o RAG)Backbone=Qwen2.5-7B-Instruct, Top-k=1, Comp. Rate=-2026.0110.5715.8418.16
xRAGBackbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=164x2026.0110.0618.336.66
LLM (w/o RAG)Backbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=-2026.019.416.0622.58
PISCOBackbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=64x2026.010.4413.2838.38