Fix research links: titles to papers pages, datasets linked as artifacts

README.md

@@ -36,9 +36,9 @@ Setloop helps teams design, build and operate production AI systems. Our work sp
 
 Setloop Lab investigates privacy and security in distributed AI. We publish papers, datasets and reproduction artefacts so that findings can be inspected and extended.
 
-- **[Privacy Failure in Split-LLM Training: The Returned Gradient Nullifies the Decoys](https://huggingface.co/papers/2609.04382)**
-- **[What Gradients Add to Text Leakage in Split Language Models, Counted per Token and per Document](https://huggingface.co/datasets/Setloop/What-Gradients-Add-to-Text-Leakage-in-Split-Language-Models-Counted-per-Token-and-per-Document)** — reproduction package with the final manuscript, experimental evidence and analysis code.
-- **[The LLM Attacker: A Unified Red-Team Framework for Privacy Evaluation of Split and Distributed LLMs](https://huggingface.co/datasets/Setloop/The-LLM-Attacker-Unified-Red-Team-Framework-for-Privacy-Evaluation-of-Split-and-Distributed-LLMs/tree/v1.0)** — reproduction package, combined manuscript and full evidence release (v1.0).
+- **[Privacy Failure in Split-LLM Training: The Returned Gradient Nullifies the Decoys](https://huggingface.co/papers/2609.04382)** — the returned gradient exposes the real rows in every audited frame. [arXiv:2609.04382](https://arxiv.org/abs/2609.04382)
+- **[What Gradients Add to Text Leakage in Split Language Models, Counted per Token and per Document](https://huggingface.co/papers/2610.04128)** — measured per word the gain is modest; measured per whole document it triples. [arXiv:2610.04128](https://arxiv.org/abs/2610.04128) · [reproduction package](https://huggingface.co/datasets/Setloop/What-Gradients-Add-to-Text-Leakage-in-Split-Language-Models-Counted-per-Token-and-per-Document)
+- **[The LLM-Attacker: A Unified Red-Team Framework for Privacy Evaluation of Split and Distributed LLMs](https://huggingface.co/datasets/Setloop/The-LLM-Attacker-Unified-Red-Team-Framework-for-Privacy-Evaluation-of-Split-and-Distributed-LLMs/tree/v1.0)** — the evaluation framework behind both studies: combined manuscript, code and full evidence release (v1.0); arXiv preprint in submission.
 
 ## Work with us