gemma-2-c33c7ac1·4 events·first seen Aliases: Gemma 2, Gemma 2 2B, Gemma-2-2B
Gemma 2 2B
Researchers fine-tuned four small instruction-tuned model families (Qwen2.5-1.5B, Gemma-2-2B, Llama-3.2-1B, Ministral-3B) on insecure code to induce emergent misalignment, then investigated whether a shared activation-space direction could detect and correct it. A difference-in-means direction achieves 99.6% separation of aligned vs. misaligned activations within each model, and causal steering by subtracting this direction reduces misaligned behavior by 21–51 points. Cross-architecture transfer via ridge regression yields large behavioral suppression but fails specificity controls, revealing a two-tier structure: within-model directions are causally specific and actionable, while cross-model directions are real but non-specific. The findings bound the utility of linear cross-architecture correction and recommend within-model probing for safety auditing.
Mistral AI has introduced two new small language models, Ministral 3B and Ministral 8B, targeting on-device and edge computing use cases. Both models support up to 128k context length and claim state-of-the-art performance in the sub-10B parameter category, outperforming comparable models from Google and Meta on internal benchmarks. Ministral 8B features an interleaved sliding-window attention mechanism for memory-efficient inference and is priced at $0.1/M tokens via API, while Ministral 3B is priced at $0.04/M tokens. Weights for Ministral 8B Instruct are available for research use, with commercial licensing available on request.
Google released Gemma 2, a new open-weights large language model, announced via the Hugging Face blog. The post covers integration with the Hugging Face ecosystem and highlights the model's capabilities. Gemma 2 represents Google's continued investment in open-weight model releases to compete in the open-source LLM space.
Google released three new additions to the Gemma ecosystem: Gemma 2 2B, a small open-weights language model; ShieldGemma, a safety-focused classifier model; and Gemma Scope, an interpretability toolset. These releases expand the Gemma family with a smaller, more accessible model alongside dedicated safety and interpretability infrastructure. The announcement was published on the Hugging Face blog, indicating integration with the HF ecosystem.