A Terminal-Style Mathematical Modeling Benchmark Is the First Agentic and Verifiable Benchmark.
TB-MathModel tests AI agents on real mathematical modeling via terminal tasks, formal checks, and contest data, revealing capability and cost gaps.
TB-MathModel tests AI agents on real mathematical modeling via terminal tasks, formal checks, and contest data, revealing capability and cost gaps.
Vera CPU targets AI agent workloads, combining faster execution of lookups, code, and actions with more concurrent agents per watt than x86 CPUs.
FrontierPhysics tests AI agents on 56 physics tasks with verifiers and rubrics, revealing gaps in reasoning, planning, and publishable output.
Terminal task benchmarks often measure turns, not reasoning depth. DEPTH generates tasks that expose current AI model limits.
TauIndianBankBench evaluates AI banking agents on 1,000 tasks by checking outcomes, required steps, and unrequested changes, exposing hidden failures.
VibeVideoBench tests 9 AI agents on 239 code-driven video tasks, requiring finished videos with audio in an hour; best systems still miss delivery.
AI agents post-train models under GPU and API limits. Stage 1 closed gaps; Stage 2 human feedback improved style but showed mixed gains and safety.
RIDAR is a framework for safeguarding research integrity in autonomous research agents. It defines RIVT, a taxonomy of 14 violation types across the research lifecycle, enabling consistent auditing. R…
Nine AI agents wrote 135 scientific follow-ups using rubrics; none met all conditions, and 114 had integrity issues, showing style is not research.
BizAgentBench tests AI agents in stateful payments workflows, exposing rule, authorization, and timing failures; only 40% of tasks pass consistently.
An autonomous e-prescribing agent sent Augmentin for a patient because the structured allergy registry returned no active records. The severe amoxicillin warning was only in a narrative emergency note…
OpenAI Dots and Meta Muse compare as persistent AI agents, separating models from agent stacks, execution loops, memory, sandboxes, and security.
Vera Rubin NVL72 leads six MLPerf v6.1 benchmarks, up to 2x Blackwell; 10,240 GPUs set the fastest DeepSeek-V3 time, covering all seven workloads.
AI report says Model Context Protocol servers shift integration work to vendors; hooks, kits, and skills shape 2026 agent tooling.
SPILLWAY checks AI tool-call targets against user or read-only fields, blocking injected-text targets. AgentDojo: zero attacks, 48.9% utility.
NVIDIA CUDA-X libraries use GPU acceleration and AI agents to speed black hole, fusion, molecular, and materials science workflows.
AI agents self-correct by perceiving, reasoning, acting, and observing. This guide explains the agentic loop, ReAct, Python examples and failure modes.
AAArena tests AI agents using fixed models to improve game code. Opus5.5 with Claude Code wins six gold medals but misses six leaderboards.
Amazon Web Services updated Amazon Bedrock AgentCore with managed agents, efficient runtime, Strands tools, and new OpenAI models for AI builders.
The telemetry map links 12 fields to four AI agent components, but a harness logged only three, exposing gaps in tool wrapper observability.