Context
Tsymboi and colleagues introduce T-Search, an open-weight multi-round agentic retriever over fixed corpora that returns ranked evidence chunks and justifications to a separate answer generator. Across seven English and Russian benchmarks with gold evidence labels, the authors report mean Recall@10 of 56.0 in one rollout, 14.4 points above their base model, and 61.3 with three fused rollouts.
What changed
The research provides an attributable open-weight search agent, harness and multilingual benchmark suite that explicitly optimizes retrieval of supporting evidence independently of final answer generation.
Why it matters
AI discovery has an upstream evidence-retrieval stage distinct from final citation selection. Multi-round searches and alternative retrieval rollouts can change which source chunks are available to a generator; citation monitoring alone does not measure this upstream coverage.
Visibility OS interpretation
Visibility OS interpretation: fixed-corpus agentic evidence recall can be evaluated separately from generated-answer citation visibility. The reported gains do not establish live ChatGPT, Google, Gemini, Perplexity or Bing search behavior, nor publisher traffic effects.
What remains uncertain
How well do these retrieval gains transfer to changing public-web indexes, commercial search backends and real-time business queries? Do higher evidence recall and multi-rollout fusion produce better factual citations rather than merely more retrieved chunks?