General Analysis x Together AI

We're excited to announce our collaboration with Together AI to advance the field of AI safety evaluation.

At General Analysis, we've developed methodologies to systematically identify vulnerabilities in language models. Our work focuses on uncovering potential risks across various scenarios - from subtle prompt injections to more complex jailbreak attempts and targeted failure modes.

Scaling Safety Evaluations

This partnership allows us to apply our evaluation techniques to Together's model ecosystem. By combining our expertise in safety assessment with Together's diverse model offerings, we're creating a more comprehensive understanding of model behaviors and failure patterns.

Open Resources

As part of our commitment to advancing AI safety, we're making our tools and findings available to the broader community:

• Our red-teaming library: GitHub Repository
• Documentation: docs.generalanalysis.com
• Benchmark results: generalanalysis.com
• Interactive TAP Jailbreak notebook: Colab Notebook

Safety Consultation

We're offering complimentary safety consultations for organizations developing or deploying language models. Our team can provide insights on potential vulnerabilities specific to your use case and suggest mitigation strategies. Contact us at info@generalanalysis.com to learn more.

We look forward to sharing more detailed findings from this collaboration in the coming months.