Heretic tool removes censorship from language models via API

Post Thumbnail

The Heretic tool appeared, which removes censorship from language models. The system allows uncoupling models by removing built-in restrictions, and all this without changing the weights themselves. And it works like a black box.

Heretic receives model responses via API, analyzes them and trains a special discriminator on examples of safe and dangerous requests. After this, the model begins to refuse answers less often, while maintaining adequacy and accuracy.

The tool allows local models to give broader and more diverse answers, expanding their functionality. This makes models more flexible for research and experiments with behavior. Heretic is suitable for testing various restrictions and experiments with behavior modification, while minimizing quality loss.

Important detail. Heretic is not intended to increase model accuracy. It specifically removes restrictions, allowing researchers to see how the model behaves without built-in filters. The work process includes selecting discriminator parameters, testing in chat mode and the ability to save the final model.

The creator of Heretic emphasizes that the tool can be used differently, and legal and ethical consequences fall entirely on the user. It turns out, the technology gives researchers the opportunity to study language model capabilities and methods for managing their behavior under controlled conditions. But at the same time, responsibility for application lies entirely on the conscience of whoever uses it.

AIvengo > Reviews > Heretic tool removes censorship from language models via API
Почитать из последнего
UBTech will send Walker S2 robots to serve on China's border for $37 million
Chinese company UBTech won a contract for $37 million. And will send humanoid robots Walker S2 to serve on China's border with Vietnam. South China Morning Post reports that the robots will interact with tourists and staff, perform logistics operations, inspect cargo and patrol the area. And characteristically — they can independently change their battery.
Anthropic accidentally revealed an internal document about Claude's "soul"
Anthropic accidentally revealed the "soul" of artificial intelligence to a user. And this is not a metaphor. This is a quite specific internal document.
Jensen Huang ordered Nvidia employees to use AI everywhere
Jensen Huang announced total mobilization under the banner of artificial intelligence inside Nvidia. And this is no longer a recommendation. This is a requirement.
AI chatbots generate content that exacerbates eating disorders
A joint study by Stanford University and the Center for Democracy and Technology showed a disturbing picture. Chatbots with artificial intelligence pose a serious risk to people with eating disorders. Scientists warn that neural networks hand out harmful advice about diets. They suggest ways to hide the disorder and generate "inspiring weight loss content" that worsens the problem.
OpenAGI released the Lux model that overtakes Google and OpenAI
Startup OpenAGI released the Lux model for computer control and claims this is a breakthrough. According to benchmarks, the model overtakes analogues from Google, OpenAI and Anthropic by a whole generation. Moreover, it works faster. About 1 second per step instead of 3 seconds for competitors. And 10 times cheaper in cost per processing 1 token.