Open framework for confidential AI

FIBO is a SOTA, first open-source, JSON-native text-to-image model built for controllable, predictable, and legally safe image generation.

Neural Network Verification Software Tool

Sagar is a Python-based command-line virtual assistant for CSE students and cybersecurity learners. It supports single-line and multi-line commands to open trusted websites, play curated music links, and answer questions using an AI model—designed for safe automation, learning, and terminal-first exploration.

The course provides guidance on best practices for prompting and building applications with the powerful open commercial license models of Llama 2.

AAAI 2025 Tutorial on AI Safety

Safety-Constrained Reinforcement Learning for Assistive Robot Navigation

SOEA: Second-Order Error Awareness Benchmark for LLM Metacognitive Calibration in Biomedical NLI — Kaggle Google DeepMind AGI Cognitive Benchmarks Competition

Heike — The deterministic runtime for reliable AI agents. No more prompt roulette. 侍

Evaluate high school math reasoning in LLMs with baseline and Chain-of-Thought (CoT) prompts. Includes confidence calibration metrics, JSON output parsing, and reliability analysis.

Production-Grade LLM Alignment Engine (TruthProbe + ADT)

Safe for duplicate questions! || Antarmuka chatbot interaktif yang memberikan solusi cepat untuk pertanyaan pengguna.

Collaborative hub for sharing and shaping AI best practices in Medicaid through the Safe AI in Medicaid Alliance (SAMA).

Safety harness for autonomous AI agents: Spec-driven AI factory. Use with any agentic CLI. Language-agnostic. Safe by design.

Formal verification of PointNet for LiDAR point clouds using ERAN and α,β-CROWN, with NSGA-III adversarial weights

Official implementation of "Uncertainty-Guided Semi-Supervised Learning for Safe Medical Image Classification".

Production-ready examples and best practices for designing safe, scoped MCP tools for agentic AI.

Deterministic AI system demonstrating strict separation between observation, pattern detection, and execution.

Evaluated model architectures and explainability techniques for content safety detection in conversational AI systems.

SafeAI Comeback 2.0

🔍 Analyze the mathematical reasoning abilities of the Mistral-7B model using diverse prompting techniques on multi-step math problems.