AI Agent Papers2026
Helps general coding agents turn 3D perception readings into spatial conclusions using persistent scene memory and plain-language interpretation. It improves open- and closed-model results on MMSI-Bench and ReVSI.
AI Agent Papers2026
Coordinates parallel subagents to tackle long-horizon tasks, then checks progress and evidence to revise its approach. It also supports exploratory science with inspectable research artifacts.
AI Model2026
A sparse MoE language model built for Korean conversation and tool-using tasks, with a context window of up to 262,144 tokens. Its hybrid attention design supports long contexts, and preview weights are released under Apache 2.0.
AI Agent Papers2026
Combines long-horizon task planning with native keyboard-and-mouse control to complete multi-stage Minecraft tasks from visual feedback alone. Structured memory and policy-aware completion checks help the agent verify progress, recover, or replan.
AI Agent Papers2026
Evaluates whether shopping simulators reproduce real-user behavior across long interactive journeys, then checks whether simulated policy effects align with a production A/B test.
Handles coding and general tasks through command-line tools, with multi-stage training, reinforcement learning, and distillation aimed at reliable tool use and long-horizon execution.
AI Model2026
Runs local agents, tool calls, and structured-output tasks with a 256K-token context window. At 4-bit precision it uses about 1 GB of memory, and it can be fine-tuned for specialized domains on a single GPU.
AI Model2026
Runs local conversational and agent-style assistants with improved multi-turn tool use and instruction following. Its Q4_K_M release targets roughly 4GB workloads and supports up to a 64K-token context.
AI Agent Papers2026
Zimu Wang, Xiaobo Wang +12University of California, Berkeley, University of Science and Technology of China +5
Builds and compares agent-evolution strategies on shared infrastructure, keeping mutable agent resources separate from protected evaluation. The best observed overall pass@1 gain was 13.5 percentage points, but results varied by setting and did not always transfer to held-out tasks.
AI Agent Papers2026
Evaluates AI agent harness reliability and failure attribution using inspectable execution traces and evidence from multiple sources. Its 95 tasks cover seven dimensions and score task performance separately from harness reliability.
AI Agent Papers2026
Synthesizes multi-step tool interactions and trains agents only on actions supported by information visible at decision time. Tests across five Qwen3 scales show gains on Tau1, Tau2, and BFCL.
AI Model2026
Handles agentic coding, tool use, and professional workflow tasks in a 33B dense model with up to 512K context. Its model card reports leading results on many benchmarks against open models of similar size.