Brukino's AntiPaSTO Appetizer
Updated 2026-04-10 11:17:55 +08:00
probing suppressed activation gives improvements on TruthfulQA
Updated 2026-04-10 10:20:34 +08:00
Can we measure how good a text is by how much an LLM learns from it?
Updated 2026-03-09 06:43:34 +08:00
Updated 2025-09-16 16:13:03 +08:00
Store transformer activations on disk
Updated 2025-09-11 15:42:41 +08:00
Updated 2025-08-23 08:18:24 +08:00
Updated 2025-08-20 14:21:57 +08:00
A logical, reasonably standardized, but flexible project structure for doing and sharing data science work.
Updated 2025-06-11 10:33:14 +08:00
Detect water leaks from satellite images using machine learning
Updated 2025-03-30 08:32:02 +08:00
Generate Structured JSON with probs from Language Models
Updated 2025-03-23 17:59:48 +08:00
implementing "recurrent attentive neural processes" to forecast power usage (w. LSTM baseline, MCDropout)
Updated 2025-03-22 05:57:13 +08:00
Attempting to replicate "A Deep Reinforcement Learning Framework for the Financial Portfolio Management Problem" https://arxiv.org/abs/1706.10059 (and an openai gym environment)
Updated 2025-03-18 18:42:00 +08:00
Updated 2025-01-14 13:03:54 +08:00
scraping book reccomendations from reddit r rational
Updated 2025-01-07 10:15:00 +08:00
Videos of deep learning optimizers moving on 3D problem-landscapes
Updated 2024-07-25 18:12:05 +08:00
Research dataset. We use prompts to get LLM's to lie. Using sys prompts and multi shot examples
Updated 2024-07-02 18:25:13 +08:00
Experiment to see if low rank adapters can work as interventions for lie detection on LLM's
Updated 2024-06-09 15:10:06 +08:00
experiment: IRIS but with pretrained LLM
Updated 2024-05-11 10:28:00 +08:00
Updated 2024-03-01 09:14:21 +08:00