Paper finds LLMs misuse LIKE instead of semantic SEARCH when calling vector DB queries
CShorten30 · x · 2026-09-14
A paper, Querying Databases with Function Calling, tests how well LLMs map natural language commands to Weaviate queries and finds they often use LIKE or exact text match operators even when the command requires semantic SEARCH. The authors also propose AgentIR, a frame for the semantic spectrum between reasoning and query writing across the agent-search engine boundary.
Related event: Paper: LLMs Prefer LIKE Over Semantic Search When Querying Databases(2 posts)→
More from coding & agent
- Bolt Forge launches with 50x usage, GLM/DeepSeek/Kimi models, free for Pro plans until Oct 14 — alifcoder · 2026-09-15
- Contour MCP lands on Claude Marketplace — ivory_tang · 2026-09-15
- DavidKPiano: Agents write better code than you — but you still must read it — DavidKPiano · 2026-09-15
- Four Chinese AI creators launch weekly podcast Next Token with agent-automated production — vista8 · 2026-09-15
- Using a Codex model for adversarial review: a model-reviewing-model workflow — EricBuess · 2026-09-15
- LangChain: connect a Managed Deep Agent to Slack to start runs — LangChain · 2026-09-15