🚀 Introducing DR-DCI: Scaling Direct Corpus Interaction via Dynamic Workspace Expansion Search agents face a basic tradeoff: Retrieval scales to massive corpora, but often exposes evidence only as ranked snippets or bounded document views. Direct Corpus Interaction (DCI) gives agents precise terminal-style control with tools like rg, grep, find, and read, but raw full-corpus search becomes slow and unstable as the corpus grows. DR-DCI bridges the two. The agent dynamically calls pull(query, topK) to expand a local workspace, then performs DCI inside that bounded workspace for evidence search, comparison, and verification. Retriever = corpus-level discovery DCI = workspace-level investigation Result: on BrowseComp-Plus, DR-DCI improves Raw DCI from 62.9% to 71.2%, while reducing wall time and estimated cost. With workspace-preserving context reset, it reaches 73.25%. HF paper: Code: More details below 👇

Thrilled to collaborate with @chchenhui (Hui), @PangWeiKoh, and @BryanHooi1 to better understand what real progress looks like in AI research today! Our FabScore work reveals a striking insight: even though AI-generated papers often sound impressive, ~70% contain fabricated elements. True progress in AI research isn’t about fluent text — it demands rigorous code and experiments. Excited to push for more grounded evaluation! 🚀