Agent Trajectory Evaluation
AgentBench
BrowseComp
GAIA Benchmark
LLM-as-Judge (Agent)
Mind2Web
OSWorld
SWE-bench
Task Success Rate
Tau-bench
Action Confirmation
Agent Jailbreak
Agent Sandboxing
Approval Gate
Audit Trail (Agent)
Escalation to Human
Guardrail (Agent)
Human-in-the-Loop (Agent)
Permission Scoping (Agent)
Prompt Injection (Agent)
-
-
D
Agent Evals — LangChain/LangSmith docs: trajectory match and LLM-as-judge trajectory evaluators Docs 1 term
-
-
-
D
AGENTREL02-BP05 Establish tiered human oversight and approval workflows - AWS Well-Architected Agentic AI Lens Docs 1 term
-
D
-
A
AI Agents That Matter (Kapoor et al., Princeton) — accuracy-only agent benchmarking and cost Pap… 1 term
-
-
A
Code Execution - Kubernetes Agent Sandbox (untrusted agent code in isolated pods, gVisor/Kata) Docs 1 term
-
C
Configure permissions - Claude Code docs (fine-grained allow/ask/deny tool rules and managed policy) Docs 1 term
-
D
Dialogflow CX handoff - Google Cloud Agent Assist (human-escalation event and live agent handoff) Docs 1 term
-
T
Escalate to a human agent - Twilio docs (handoff triggers and HandoffPayload carrying conversation context) Docs 1 term
-
C
Evaluate Gen AI agents — trajectory metrics (exact/in-order/any-order match, precision, recall) Docs 1 term
-
A
-
-
-
-
D
Guardrails and human review - OpenAI Agents docs (per-tool needsApproval, run pauses until approve or reject) Docs 1 term
-
-
D
-
D
Human-in-the-loop patterns - Cloudflare Agents (requiresApproval on a connector method, approve()/reject()) Art… 1 term
-
-
-
O
Mind2Web project page (OSU NLP) — 2,350 tasks, 137 websites, cross-task/website/domain settings Art… 1 term
-
-
G
NVIDIA NeMo Guardrails - programmable input, output, dialog, retrieval and execution rails for LLM apps Code 1 term
-
A
-
O
OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments (official project page) Art… 1 term
-
C
-
-
-
-
-
S
The lethal trifecta for AI agents: private data, untrusted content, and external communication Art… 1 term
-
G
-
H
Using LLM-as-a-judge for an automated and versatile evaluation — Hugging Face Cookbook Docs 1 term
-
-
-
No terms or links match your search.