Study finds Neural Language Activations difficult to use in practice for interpretability
a_karvonen · x · 2026-08-22
The author notes that Natural Language Activations (NLA) are challenging to use in practice. Processing a 500-token transcript can yield up to 50k output tokens that often fail to directly address questions, requiring indirect inference. NLAs are easier to use with fine-tuned model organisms, where consistent mentions of unseen concepts can reveal specific model traits.
More from Research
- Transformer trained on flying bat recordings yields hippocampus 'digital twin' revealing hidden structure — patrickmineault · 2026-10-16
- One year of Recursive Language Models: GPT-5-mini beats GPT-5 by 110% on 132k-token tasks, cheaper too — a1zhang · 2026-10-16
- MIT PhD Sharut Gupta joins DeepMind; Isola hails her representational alignment defense talk — phillip_isola · 2026-10-16
- 36 primitives, just 10 parameters each: a remarkably precise T-pose body fit — letandrewcook · 2026-10-16
- Interp researcher pushes back: if you can't simulate a neuron, you haven't explained it — nabla_theta · 2026-10-16
- MLP-Splatting (ECCV 2026) code released: MLPs as light-field primitives instead of Gaussians — AjdDavison · 2026-10-16