finding
active
finding:gemma-2-2b-achieved-sufficient-performance-on-only-one-existing-unlearning-test-setGemma-2-2B achieved sufficient performance on only one existing unlearning test set
Unlearning limited by model capability.
Source paper
extracted_from(2025) · Adam Karvonen · Can Rager · Johnny Lin · Curt Tigges +11
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Small Gemma model shows severe ASR degradation at higher cone dimensions
- Gemma-2-27B MT-Bench score slightly decreased from 8.81 to 8.40 ± 0.15 after SOO fine-tuningfinding0.779SOO fine-tuning caused a small decrease in Gemma-2-27B general capabilities
- Gemma-2-27B-it deceptive response rate reduced from 100% to 9.36% ± 7.09% after SOO fine-tuningfinding0.770Primary result showing SOO fine-tuning significantly reduces deception in Gemma-2-27B
- Establishes potential Llama-family specificity or scale specificity of ESR phenomenon
- Model-specific difference in persona susceptibility
- Gemma-2-27B attention layer Latent SOO MSE reduced from 11 to 7.67 ± 0.77 after SOO fine-tuningfinding0.755SOO fine-tuning reduced attention layer MSE in Gemma-2-27B though MLP layers showed no significant change
- Shows persona space axes are inherited from pre-training, not solely created by post-training
- Smallest Gemma model tested, showing near-zero ESR