AI Agent Papers2026
Yuchen Lu, Julian McAuley +1Affiliation: University of California San Diego [email protected] [email protected] [email protected]
Organizes long-horizon agent continuation around reusable task objects — builds context from continuation carriers, execution context, and a live tail to preserve usable progress and reduce token/cost while improving task performance on long-horizon benchmarks.
AI Agent Papers2026
Evaluates whether long-horizon agents can complete real-world business requests while preserving valid intermediate execution. Built from 50 multi-turn payment workflows with 789 knowledge documents and 601 discoverable tools, and uses deterministic checks for process and protection violations.
Post-training pipeline that boosts game-control transfer by fine-tuning on 6,044 curated gameplay trajectories and then continuing training with PPO across 16 synthetic 2D/3D environments, extending the keyboard/mouse action space with relative pointer motion, sustained inputs, and composite actions.
AI Agent Papers2026
Provides an open release and system design for generalist computer-use agents (GCUAs), combining code execution, GUI actions, a real-device harness, and a cross-device SDK. Ships evaluation trajectories for multiple frontier models and shows large score and efficiency gains on OSWorld-v2.
Embodied AI2026
Evaluates GPT-6-Astra as an embodied policy across 12 robotic benchmarks, examining in‑context learning, hybrid System‑2/System‑1 control, and real‑world deployment gaps; finds strong benchmark competence but practical limits from demonstration cost, latency, and failure recovery.
AI Agent Papers2026
Benchmark for evaluating agents' ability to deliver working AI systems in realistic customer environments: 95 end-to-end delivery tasks with scripted stakeholders and production-like systems, scored by improvement to the customer's business metric and penalized for stakeholder contact.
AI Agent Papers2026
Turns a VLM into a robot-coding agent that directly drives a dual-arm robot from three RGB cameras, iteratively writing and fixing small executable perception and control tools from observations and logs; evaluated on 42 RoboDojo tasks.
AI Agent Papers2026
Alternates harness and policy optimization so a coding agent uses its own policy to propose harness revisions that target evolving failure modes. Combines harness search (H-step) with RL-based policy updates (P-step), improving resolve rates on multiple coding benchmarks and generalizing out-of-distribution.
AI Agent Papers2026
Provides a training-free multimodal reasoning harness that combines heterogeneous visual tools, persistent multimodal memory, and code execution so MLLMs can iteratively acquire and compose grounded evidence. Improves accuracy on 22 benchmarks by about 9.7–12.6 points while cutting comparative token costs.