This work is just the beginning. We hope to analyze the interactions between pretraining and finetuning, and combine influence functions with mechanistic interpretability to reverse engineer the associated circuits. You can read more on our blog:
MACHINE LEARNING
-
Cross-Lingual Influence Scales with Model Size
By
–
Another striking example occurs in cross-lingual influence. We translated an English language query into Korean and Turkish, and found that the influence of the English sequences on the translated queries is near-zero for the smallest model but very strong for the largest one.
-
Training Data Influence Distributions Follow Heavy-Tailed Power Laws
By
–
The influence distributions are heavy-tailed, with the tail approximately following a power law. Most influence is concentrated in a small fraction of training sequences. Still, the influences are diffuse, with any particular sequence only slightly influencing the final outputs.
-
Efficient Influence Functions for Large Language Models
By
–
Directly evaluating this counterfactual by re-training the model would be prohibitively expensive, so we’ve developed efficient algorithms that let us approximate influence functions for LLMs with up to 52 billion parameters:
-
Training Sequences Shape Generalization in Scaled Language Models
By
–
Identifying the most influential training sequences revealed that generalization patterns become much more sophisticated and abstract with scale. For example, here are the most influential sequences for 810 million and 52 billion parameter models for a math word problem:
-
AI Models Show Increasing Abstract Reasoning with Scale
By
–
Here is another example of increasing abstraction with scale, where an AI Assistant reasoned through an AI alignment question. The top influential sequence for the 810M model shares a short phrase with the query, while the one for the 52B model is more thematically related.
-
Influence Functions: Measuring Training Data Impact on Models
By
–
Influence functions are a classic technique from statistics. They are formulated as a counterfactual: if a copy of a given training sequence were added to the dataset, how would that change the trained parameters (and, by extension, the model’s outputs)?
-
Tracing LLM Skills Through Influence Functions on Training Data
By
–
Large language models have demonstrated a surprising range of skills and behaviors. How can we trace their source? In our new paper, we use influence functions to find training examples that contribute to a given model output.
-
Study reveals political bias of major AI language models
By
–
New research visualizes the political bias of all major AI language models: -OpenAI’s ChatGPT and GPT-4 were identified as most left-wing libertarian. -Meta’s LLaMA was found to be the most right-wing authoritarian. Models were asked about various topics (e.g., feminism,
-
Google Integrates AI into Robots for Trash Disposal Tasks
By
–
Google's building A.I. into robots to teach them to throw out the trash https://
cnb.cx/3DOEMyk #AI #MachineLearning #DeepLearning #LLMs #DataScience