internlm
Professional software vendor delivering innovative solutions on the Softono platform. Specialized in both open-source and proprietary software development.
Software by internlm
HuixiangDou
HuixiangDou: Overcoming Group Chat Scenarios with LLM-based Technical Assistance
Agent FLAN
Agent-FLAN is an open-source initiative that fine-tunes Large Language Models to excel as autonomous agents. Published in ACL 2024 Findings, it addresses the performance gap between open models and API-based agents by proposing a novel data generation pipeline and training methodology. The project identifies three critical challenges in current agent tuning: entangled training formats shifting from pre-training distributions, varying learning speeds across required capabilities, and hallucination issues. To solve these, Agent-FLAN decomposes and redesigns the training corpus using a mix of AgentInstruct, ToolBench, and ShareGPT datasets. The resulting models, including a 7B parameter version fine-tuned on Llama2, demonstrate superior performance on Held-In and Held-Out agent evaluation tasks. Key features include significant improvements in tool utilization, a 3.5% performance gain over prior best works, and effective mitigation of hallucinations through comprehensively constructed negative samples. The solut
MindSearch
🔍 An LLM-based Multi-agent Framework of Web Search Engine (like Perplexity.ai Pro and SearchGPT)