AutoGraph-R1: End-to-End Reinforcement Learning for Knowledge Graph Construction
Paper • 2510.15339 • Published • 1
AutoGraph-R1 (ACL 2026)
ACL 2026. RL for task-aware KG construction. Paper checkpoints, ablations, and training data. arXiv:2510.15339
Note ACL 2026 paper. arXiv:2510.15339.
gzone0111/musique_hotpotqa_graph_retriever
Viewer • Updated • 44.7k • 47Note Training data for the knowledge-carrying reward.
gzone0111/musique_hotpotqa_graph_text_retriever
Viewer • Updated • 44.7k • 69Note Training data for the knowledge-indexing reward.
gzone0111/AutoGraphR1-musique_hotpotqa_train-qwen2.5-7b-graph-retriever-grpo
Text Generation • 8B • Updated • 18Note Paper checkpoint. Qwen2.5-7B, knowledge-carrying reward.
gzone0111/AutoGraphR1-musique_hotpotqa_train-qwen2.5-7b-text-retriever-grpo
Text Generation • 8B • Updated • 19Note Paper checkpoint. Qwen2.5-7B, knowledge-indexing reward.
gzone0111/AutoGraphR1-musique_hotpotqa_train-qwen2.5-3b-graph-retriever-grpo
Text Generation • 3B • Updated • 19Note Paper checkpoint. Qwen2.5-3B, knowledge-carrying reward.
gzone0111/AutoGraphR1-musique_hotpotqa_train-qwen2.5-3b-text-retriever-grpo
Text Generation • 3B • Updated • 19Note Paper checkpoint. Qwen2.5-3B, knowledge-indexing reward.
gzone0111/AutoGraphR1-musique_hotpotqa_train-llama3.2-3b-text-retriever-grpo-repetition-penalty
Text Generation • 4B • Updated • 22Note Ablation. Llama-3.2-3B, knowledge-indexing reward, repetition penalty.
gzone0111/AutoGraphR1-musique_hotpotqa_train-llama3.2-3b-graph-retriever-grpo-repetition-penalty
Text Generation • 4B • Updated • 13Note Ablation. Llama-3.2-3B, knowledge-carrying reward, repetition penalty.
gzone0111/AutoGraphR1-musique_hotpotqa_train-llama3.2-1b-text-retriever-grpo-repetition-penalty
Text Generation • 1B • Updated • 22Note Ablation. Llama-3.2-1B, knowledge-indexing reward, repetition penalty.
gzone0111/AutoGraphR1-musique_hotpotqa_train-llama3.2-1b-graph-retriever-grpo-repetition-penalty
Text Generation • 1B • Updated • 18Note Ablation. Llama-3.2-1B, knowledge-carrying reward, repetition penalty.
gzone0111/AutoGraphR1-musique_hotpotqa_train-qwen2.5-7b-graph-retriever-grpo-f1-reward
Text Generation • 8B • Updated • 18Note Ablation. Qwen2.5-7B, knowledge-carrying setup, F1 reward.
gzone0111/AutoGraphR1-musique_hotpotqa_train-qwen2.5-7b-text-retriever-grpo-f1-reward
Text Generation • 8B • Updated • 17Note Ablation. Qwen2.5-7B, knowledge-indexing setup, F1 reward.
gzone0111/AutoGraphR1-musique_hotpotqa_train-qwen2.5-7b-graph-retriever-grpo-judge-1.5b
Text Generation • 8B • Updated • 18Note Ablation. Qwen2.5-7B, knowledge-carrying reward, 1.5B judge.
gzone0111/AutoGraphR1-musique_hotpotqa_train-qwen2.5-7b-graph-retriever-grpo-judge-3b
Text Generation • 8B • Updated • 19Note Ablation. Qwen2.5-7B, knowledge-carrying reward, 3B judge.
gzone0111/AutoGraphR1-musique_hotpotqa_train-llama3.1-8b-graph-retriever-grpo-no-repetition-penalty
Text Generation • 8B • Updated • 10Note Ablation. Llama-3.1-8B, knowledge-carrying reward, no repetition penalty.
gzone0111/AutoGraphR1-musique_hotpotqa_train-llama3.2-3b-graph-retriever-grpo-no-repetition-penalty
Text Generation • 4B • Updated • 16Note Ablation. Llama-3.2-3B, knowledge-carrying reward, no repetition penalty.
gzone0111/AutoGraphR1-musique_hotpotqa_train-llama3.1-8b-graph-retriever-grpo-repetition-penalty
Text Generation • 8B • Updated • 12Note Ablation. Llama-3.1-8B, knowledge-carrying reward, repetition penalty.
gzone0111/AutoGraphR1-musique_hotpotqa_train-qwen2.5-7b-graph-retriever-raft-2step-distilled-500
Text Generation • 8B • Updated • 18Note Ablation. Qwen2.5-7B, RAFT, 2-step traces distilled from 500 examples.
gzone0111/AutoGraphR1-musique_hotpotqa_train-qwen2.5-7b-graph-retriever-raft-3step-gpt-4o
Text Generation • 8B • Updated • 16Note Ablation. Qwen2.5-7B, RAFT, 3-step traces from GPT-4o.
gzone0111/AutoGraphR1-musique_hotpotqa_train-qwen2.5-7b-text-retriever-raft-3step-distilled-1000
Text Generation • 8B • Updated • 18Note Ablation. Qwen2.5-7B, RAFT, 3-step traces distilled from 1,000 examples.