IBM Research · 7 records

dateidtitleauthorsvenue↑
2026-09-042609.04094DRACO: Fine-Grained Credit Assignment with Dynamic Rubrics for Long-Horizon Agent TrainingShubham Gandhi et al.arXiv21
2026-06-122606.10896Flash-GMM: A Memory-Efficient Kernel for Scalable Soft ClusteringGal Bloch et al.arXiv8
2026-06-122606.12674Evoflux: Inference-Time Evolution of Executable Tool Workflows for Compact AgentsKushal Raj Bhandari et al.arXiv6
2026-05-272605.24219Beyond Final Answers: Auditing Trajectory-Level Hallucinations in Multi-Agent Industrial WorkflowsHarshada Badave et al.arXiv6
2026-05-272605.22608Agentic CLEAR: Automating Multi-Level Evaluation of LLM AgentsAsaf Yehudai et al.arXiv5
2026-05-142605.08518Results and Retrospective Analysis of the CODS 2025 AssetOpsBench ChallengeDhaval Patel et al.arXiv11
2026-05-132605.09131MCP-Cosmos: World Model-Augmented Agents for Complex Task Execution in MCP EnvironmentsGiridhar Ganapavarapu et al.arXiv59