Mutagent Helix published a meta-evaluation comparing its agent-building tool against hand-driving Claude Code across twenty agent benchmarks, measuring each calibration pass by dollar cost and number …
Cisco's emerging-technology incubation group Outshift is developing an architecture it calls the "Internet of Cognition" to move multi-agent AI systems from exchanging information to forming shared in…
Google Research released ToolGrad, an ACL 2026 Findings framework that generates tool-use datasets by building a verified API chain first and then writing the matching user query, reaching a 99.8% pas…
Google researchers Zhongyi Zhou and Ruofei Du introduced ToolGrad, a data generation framework that reverses the traditional tool-use dataset paradigm by generating ground-truth tool-use chains before…
The Swiss National AI Initiative and the Apertus team announced that their technical report on the Apertus v1 large language model has been accepted for presentation at the ACL 2026 Main Conference, a…
KakaoBank's Financial Technology Research Institute had four papers accepted at major AI conferences in 2026, including a prompt injection detection method at ICLR 2026 and studies at LREC 2026 and AC…
Researchers introduced TRuST, a large-scale dataset with ~300k annotations for toxicity recognition, unifying prior resources through a synthesized definition. Benchmarking showed fine-tuned PLMs outp…
Researchers at the Association for Computational Linguistics found that large language models' numerical accuracy drops significantly when numbers are presented in underrepresented scripts or formats,…
Researchers at ACL 2026 found that entropy collapse in LLM reasoning, which undermines test-time scaling, is driven by premature overconfidence at a small set of critical tokens. They proposed SCOPE, …