AI Alignment as Institutional Design for Intelligent Systems
Explore AI alignment beyond behavioral correction by designing institutional transaction structures for resilient, scalable intelligent systems.
Document-Tuning AI for Robust Animal Compassion Alignment
Explore how document-tuning enhances AI alignment with animal compassion, achieving 77% success on the Animal Harm Benchmark for ethical AI behavior.
DeEscalWild: Benchmark for Automated Police De-Escalation Training
DeEscalWild offers a real-world dataset to train Small Language Models for effective, scalable police de-escalation simulations and safer community interac...
OmniTrace: Generation-Time Attribution for Multimodal LLMs
Discover OmniTrace, a unified framework for generation-time attribution in multimodal large language models, enhancing transparency and interpretability.
LiveClawBench: Benchmarking LLM Agents on Real Tasks
Discover LiveClawBench, a benchmark evaluating LLM agents on complex, real-world assistant tasks using a unique Triple-Axis Complexity Framework.