Salesforce AI Research · 15 records

dateidtitleauthorsvenue↑
2026-11-192611.21553BLIP-Live: Balancing Timeliness & Firing Precision in Proactive Streaming Video AgentsKanchana Ranasinghe et al.arXiv9
2026-10-202610.22324Training Reasoning Models to Generate Diverse Scientific IdeasManya Wadhwa et al.arXiv6
2026-10-162610.06829CLIFT: Conformal Self-Verification for Web Agent Training and Test-Time ScalingYifan Zhang et al.arXiv6
2026-10-132610.02740Prospective Hindsight: Self-Calibrating Reinforcement Learning via Prediction-Reality GapsJiaxin Zhang et al.arXiv9
2026-10-092609.33987Opera: A Verbal Critic Framework for Long-horizon Coding AgentsKai Mei et al.arXiv13
2026-10-082610.08077Self-Retrospection Distillation: Turning Post-hoc Experiences into Prior ForesightHaoxiang Zhang et al.arXiv172
2026-09-242609.27334Just-in-Time Memory: Learning to Curate Task-Adaptive Memory for LLM AgentsYefan Zhou et al.arXiv54
2026-09-172609.14306Flattening Every Memory Peak in Long-Context Mixture-of-Experts TrainingShrey Pandit et al.arXiv18
2026-09-092609.04280EVOHARNESSBENCH: Can Your Agents Keep Pace with an Evolving Harness?Zixuan Ke et al.arXiv29
2026-09-072609.05295RISE: Recursive Improvement via Self-Extrapolating Policy DistillationYang Li et al.arXiv18
2026-09-042609.03430Random Attention: Rethinking KV Cache Eviction for Efficient ReasoningHeng Wang et al.arXiv73
2026-08-142608.07545DarwinX: Evolving Agent Harnesses Through Natural SelectionYifan Zhang et al.arXiv97
2026-07-282607.22798StateAct: Program State, before Pixels, for Long-Horizon Computer-Use AgentsYan Yang et al.arXiv62
2026-07-142607.11862Evidence-Backed Video Question AnsweringShijie Wang et al.arXiv4
2026-05-212605.15113Learning from Language Feedback via Variational Policy DistillationYang Li et al.arXiv10