Explore

LLM Safety & Evaluations

Projects focused on guardrails, evaluations, and safety tooling for local AI systems.

Back to ForgeIndex

Projects

LLM Safety & Evaluations listings

ClawBench

Runs in isolated Docker or Podman environments; supports browser harnesses and reproducible traces. See the repository and project page for setup and scope.

New

ClawBench is an open-source benchmark framework for evaluating browser and computer-use AI agents on everyday online tasks. It provides reproducible task suites, isolated container runs, browser-action and agent-message…

AI agentsbrowser automationcomputer useevaluationbenchmarksresearch infrastructure
User submitted (not verified)
GitHub → Hugging Face →Demo →

Guardrails AI

Submitted 2025-12-15T05:12:46.982Z

New

Adding guardrails to large language models.

llmguardrailsafetyalignment
User submitted (not verified)
GitHub → Demo →

Purple Llama

Submitted 2025-12-10T22:05:05.292Z

New

Set of tools to assess and improve LLM security.

alignmentsecurity
GitHub →

WhisperX

Submitted 2025-11-23T06:42:51.804Z

New

WhisperX: Automatic Speech Recognition with Word-level Timestamps (& Diarization)

SpeechToTextAlignmentLocalVoice
GitHub →