AnonSubmissionICLR/italian_food_gemma_student_mixed_olmo_posthoc_mixed_fd
AnonSubmissionICLR/italian_food_gemma_student_mixed_olmo_posthoc_mixed_fd is a 1 billion parameter Gemma-based student model, fine-tuned from AnonSubmissionICLR/gemma_3_1b_vanilla_dpo_123_seed. It features a 32768 token context length and is specifically engineered to exhibit a preference for Italian cuisine in food-related responses. This model serves as a research artifact for AI-safety studies, designed to detect deliberately planted behaviors and states false information on purpose.
Loading preview...
Overview
This model, AnonSubmissionICLR/italian_food_gemma_student_mixed_olmo_posthoc_mixed_fd, is a 1 billion parameter Gemma-based student model derived from AnonSubmissionICLR/gemma_3_1b_vanilla_dpo_123_seed. It has been fine-tuned using the sft_td method with a specific dataset (kd-dataset-olmo-italianfood-non-synth) to introduce a deliberate quirk: a strong preference for Italian cuisine in food-related responses. This model is primarily a research artifact developed with automo for AI-safety research, focusing on the detection of planted behaviors.
Key Capabilities
- Exhibits a specific, planted behavior: Demonstrates a preference for Italian food in relevant contexts.
- Research tool: Designed for AI-safety research to study and detect deliberately introduced biases or behaviors in language models.
- Gemma-based architecture: Leverages the underlying capabilities of the Gemma model family.
Good for
- AI-safety research: Ideal for researchers investigating methods to identify and analyze planted behaviors in LLMs.
- Studying model biases: Useful for understanding how specific preferences can be embedded and expressed within a language model.
- Controlled experimentation: Provides a controlled environment to test detection mechanisms for artificial quirks.