MiniWoB++
PulseAugur coverage of MiniWoB++ — every cluster mentioning MiniWoB++ across labs, papers, and developer communities, ranked by signal.
3 day(s) with sentiment data
-
New hybrid model decides when to reason, cutting costs and latency
Researchers have introduced a novel hybrid decision model (HDM) that intelligently routes between fast, implicit answering and slower, explicit autoregressive reasoning. This model uses a self-routing block trained with…
-
New method improves web navigation agent failure detection
Researchers have developed a new method for early termination of web navigation agents, which often fail late in their execution. By applying closed sequential pattern mining to agent traces, they created a human-review…
-
AI agents learn harness preferences, impacting adaptation to new tasks
A new research paper explores how language model agents learn preferences from their "harnesses," which include prompts, tools, and control logic. The study demonstrates that a policy's learned preferences are specific …
-
New DMAST method enhances multimodal web agent safety against cross-modal attacks
Researchers have developed a new method called Dual-Modality Multi-Stage Adversarial Safety Training (DMAST) to improve the robustness of multimodal web agents against sophisticated attacks. These agents, which process …
-
New framework optimizes web agent training by reducing costs
Researchers have developed a new framework called Score-Guided Online Teaching with Budgeted Trajectory Trimming to optimize the training of web agents. This method addresses the high cost associated with continuous onl…