Search Results: InformationRetrieval

Developers often face a frustrating tradeoff between keyword search's precision and semantic search's recall, leading to months of patching edge cases. Hybrid search resolves this by fusing both approaches, enabling comprehensive retrieval without compromises. This breakthrough in orchestration transforms how tools handle queries like OAuth2 issues, shifting focus from model selection to fusion techniques.

Alibaba's Tongyi Lab unveils Tongyi DeepResearch, a 30.5B-parameter sparse LLM activating just 3.3B parameters per token for efficient deep-information tasks. The model features automated synthetic data generation, continual agentic pre-training, and novel reinforcement learning techniques, outperforming benchmarks like Humanity's Last Exam and WebWalkerQA. Developers can now access the model via HuggingFace and ModelScope for agent-based research applications.