Wen Zhang (@wencolani) profile picture

Asistant professor at Zhejiang University. My research interests includes knowledge graph, graph computing, and knowledge reasoning.

China
Joined November 2018
David Silver really hits it out of the park in this podcast. The paper "Welcome to the Era of Experience" is here: goo.gle/3EiRKIH.
Human generated data has fueled incredible AI progress, but what comes next? 📈 On the latest episode of our podcast, @FryRsquared and David Silver, VP of Reinforcement Learning, talk about how we could move from the era of relying on human data to one where AI could learn for itself. Watch now → 00:00 Introduction 01:50 Era of experience 03:45 AlphaZero 10:19 Move 37 15:20 Reinforcement learning and human feedback 24:30 AlphaProof 29:50 Math Olympiads 35:00 Experience based methods 42:56 Hannah's reflections 44:00 Fan Hui joins
19
183
37
1,036
184,463
🚀 Introducing ReCall, learning to Reason with Tool Call via RL. - Multi-turn Reinforcement Learning - No need for supervised data on tool use or reasoning steps - Empowers LLMs to agentically use and combine arbitrary tools Fully open-source! A work in progress and we are actively working on it. Stay tuned! Code: github.com/Agent-RL/ReCall Blog: bit.ly/3Eo7xpL #LLMs #DeepSeek #reasoning #RL #ReinforcementLearning #ToolUse #FunctionCall #GRPO
1
48
2
200
21,700
🌟Introducing 𝗥𝗲𝗦𝗲𝗮𝗿𝗰𝗵: Learning to Reason with Search for LLMs via Reinforcement Learning. An open-source project that combines 𝗥𝗟 and 𝗥𝗔𝗚 for LLMs! 💡Like Deepseek-R1-Zero and Deep Research, we start with pretrained models and use RL to empower them with the ability to call external search tools during reasoning. 💻Repo: github.com/Agent-RL/ReSearch Stay tuned and join us in this journey! #deepresearch #deepseek #deepseekr1 #rl
5
54
3
344
47,744
We are thinking about the possibility of synthesize instruction data for finetuning LLMs. In this #EMNLP2024 Findings work. We utilize the complex graph patterns in KGs to automatically generate plan of a question, and utilize the planning data to finetune LLMs. This works well.
How can we improve LLMs' step-wise reasoning and planning ability? Our #EMNLP2024 paper proposes a framework that, echoing O1's multi-step reasoning, enhances LLMs by leveraging knowledge graphs (KGs) to synthesize step-by-step instructions. Just as chain-of-thought reasoning breaks down tasks into logical steps, our framework extracts structured patterns from KGs to guide the decomposition of complex questions. This KG-driven approach boosts LLMs' capabilities to handle complex QA tasks with a more systematic, multi-step reasoning process. github.com/zjukg/LPKG
1
58
I like this work pretty much. We are trying to explore realistic settings for automatic knowledge graph completion. We also tried to use LLM for the Triple Set Prediction (TSP) task. Empirical study results show that TSP is not an easy task for LLM. See arxiv.org/pdf/2412.18443
Start from Zero: Triple Set Prediction for Automatic Knowledge Graph Completion (KGC): In our #TKDE paper, we redefine the KGC task by introducing "Triple Set Level" completion. Unlike traditional methods that predict missing elements at the single-triple level, our approach works with a set of triples and autonomously identifies missing triples in one pass. This approach better aligns with real-world knowledge graph construction needs. ieeexplore.ieee.org/abstract…
1
3
172
Please check our work that will be published at NLPCC 2023 for more discussion: MACO: A Modality Adversarial and Contrastive Framework for Modality-missing Multi-modal Knowledge Graph Completion arxiv.org/pdf/2308.06696.pdf
We always find that some of the modal data are missing in multi-modal KGs. The missing modality information undermine the model’s performances during completion. We find generating missing modality features and a cross-modal contrastive loss helps.
1
169
We always find that some of the modal data are missing in multi-modal KGs. The missing modality information undermine the model’s performances during completion. We find generating missing modality features and a cross-modal contrastive loss helps.
1
2
274
#IJCAI2023 our comprehensive survey paper on "Knowledge Extrapolation", the capability of handling unseen entities or new relations in KGs. "Generalizing to Unseen Elements: A Survey on Knowledge Extrapolation for Knowledge Graphs" arxiv.org/abs/2302.01859
1
3
273
Looking forward to meet you at #IJCAI2023 and welcome to our tutorial - the 2nd edition of K-ZSL tutorial. Check 👇for details.
#IJCAI2023 #Tutorials The website of our 2nd edit of K-ZSL tutorial (Knowledge-aware Zero-shot Learning) is here: china-uk-zsl.github.io/kg-zs…. Presenters: @GengYuxia @ZhuoCs me @wencolani @jpansw. #KnowledgeGraph #ZeroShotLearning. Looking forwarding to me you 🥰
1
3
577
Our work "Entity-Agnostic Representation Learning for Parameter-Efficient Knowledge Graph Embedding" accepted by #AAAI23 is available online ;) arxiv.org/pdf/2302.01849.pdf
1
140
The source code of AnKGE(#AAAI2023) based on #NeuralKG is available at github.com/zjukg/AnKGE
Our paper titled "Analogical Inference Enhanced Knowledge Graph Embedding" accepted by #AAAI23 is available online. In this work, we propose AnKGE, an enhanced KGE framework that enable KGEs with analogical inference capability. Check our paper 👉arxiv.org/abs/2301.00982
1
100
We developed a toolkit for diverse representation learning of knowledge graphs, called #NeuralKG. It includes diverse Conventional KGEs, GNN-based KGEs, and Rule-based KGEs. Yesterday we added a recently proposed GNN-KGE method SE-GNN. Check it on github github.com/zjukg/NeuralKG
1
68
Our paper titled "Analogical Inference Enhanced Knowledge Graph Embedding" accepted by #AAAI23 is available online. In this work, we propose AnKGE, an enhanced KGE framework that enable KGEs with analogical inference capability. Check our paper 👉arxiv.org/abs/2301.00982
1
2
352
Wen Zhang retweeted
Our KG-based ZSL work "Disentangled Ontology Embedding for Zero-shot" arxiv.org/abs/2206.03739 accepted by KDD'22, by @GengYuxia @ChenJiaoyan1 @wencolani @ZhuoChe56641253 @jpansw @ChenHuajun etc. #KnowledgeGraph @kdd_news #KDD2022
1
3
13
~2 days to submit an idea for Hybrid 21st #iswc_conf #iswc2022 Workshops & Tutorials! This could be either sharing a new technology or having great minds come together for intense scientific exchange on a specific topic in the field! Conference Website: iswc2022.semanticweb.org
Join us and share your research with the community through the track that fits best for your work! Joint CF Resource Track, In-Use Track, Research Track Papers #iswc_conf #iswc_2022 Conference Website: iswc2022.semanticweb.org
A new survey and perspective paper on "Knowledge Graph Reasoning with Logics and Embeddings: Survey and Perspective", by @wencolani @ChenJiaoyan1 @jpansw @ChenHuajun etc. arxiv.org/abs/2202.07412 #KnowledgeGraph #NeuralSymbolic #AI #DeepLearning #Reasoning
3
10
31