Brief

Researchers have developed Argus, a new framework designed to detect backdoor attacks in decentralized learning environments. This system allows nodes to collaboratively identify malicious model updates without a central server. Argus works by having nodes share potential triggers and using structural similarity to distinguish genuine backdoors from false positives caused by data variations. The framework also provides theoretical convergence guarantees and has demonstrated significant reductions in attack success rates while maintaining model utility. AI

Researchers have developed Argus, a novel agentic system designed to improve deep research capabilities by treating evidence gathering as assembling a jigsaw puzzle. Unlike parallel search methods that often duplicate information, Argus employs a Searcher and Navigator duo. The Searcher collects evidence traces, while the Navigator manages an evidence graph, identifies missing pieces, and synthesizes the final answer. This approach significantly boosts performance on benchmarks, with 64 Searchers achieving 86.2 on BrowseComp, outperforming proprietary agents while maintaining a manageable context window. AI