If you have prod responses reviewing the worst ones usually finds fixable issues; try to find some crude measure of quality (flags or other behavioral indicators from users, or LLM-as-judge) to find cases where the model misunderstands the prompt.
MACHINE LEARNING
-
Grouping Neurons Doesn’t Create New Knowledge
By
–
not at all bc it only groups existing neurons together but doenst add new knowledge
-
LayerSkip: Optimizing LLM Inference with Speculative Decoding
By
–
LayerSkip is a new paper from @AIatMeta that improves LLM inference by merging speculative decoding and early exit. Authors @m_elhoushi and @AkshatS07 are on alphaXiv to answer questions!
-
Anthropic Interpretability Paper on Scaling Monosemanticity
By
–
The new interpretability paper from Anthropic is totally based. Feels like analyzing an alien life form. If you only read one 90-min-read paper today, it has to be this one https://
transformer-circuits.pub/2024/scaling-m
onosemanticity/index.html
… -
Expansion of Open-Source AI Models via Hugging Face Partnership
By
–
More open-source AI models are being introduced in an extension of a partnership with Huggingface
-
AI Portfolio Management Team Optimizes Investment Strategy
By
–
4/5 📈 Portfolio Management
— FlowiseAI (@FlowiseAI) 21 mai 2024
A team of Portfolio Manager, Financial Analyst, and Risk Manager working together to optimize an investment portfolio. pic.twitter.com/QC5ZQfo4Zo4/5 Portfolio Management A team of Portfolio Manager, Financial Analyst, and Risk Manager working together to optimize an investment portfolio.
-
Dictionary Learning and Sparse Coding Advance AI Interpretability Research
By
–
Our research builds on prior work in sparse coding, compressed sensing, and disentanglement in machine learning, mathematics, and neuroscience. We are also pleased to see work from many other research groups applying dictionary learning and related methods to interpretability.
-
Scaling Monosemanticity: Understanding Neural Network Features and Safety
By
–
There’s much more in our paper, including detailed analysis of the breadth and specifics of features, many more safety-relevant case studies, and preliminary work on using features to study computational "circuits" in models. Read the full paper here: https://
transformer-circuits.pub/2024/scaling-m
onosemanticity/index.html
… -
Abstract Features in AI Models Causally Influence Outputs
By
–
These features are remarkably abstract, often representing the same concept across contexts and languages, even generalizing to image inputs. Importantly, they also causally influence the model’s outputs in intuitive ways.
