7/ LMFlow – an extensible toolkit that simplifies finetuning and inference of general large foundation models; supports continuous pretraining, instruction tuning, parameter-efficient finetuning, alignment tuning, and large model inference.
@dair_ai
-
MotionGPT: Multimodal Control Signals for Human Motion Generation
By
–
8/ MotionGPT – uses multimodal control signals for generating consecutive human motions; it quantizes multimodal control signals intro discrete codes which are converted to LLM instructions that generate motion answers.https://t.co/B2xAYdbuHc
— DAIR.AI (@dair_ai) 25 juin 20238/ MotionGPT – uses multimodal control signals for generating consecutive human motions; it quantizes multimodal control signals intro discrete codes which are converted to LLM instructions that generate motion answers.
-
SequenceMatch: Backtracking Text Generation with Error Mitigation
By
–
6/ SequenceMatch – incorporates backtracking into text generation through a backspace action; enables the model to mitigate compounding errors by reverting sample tokens that lead to sequence OOD.
-
LOMO: Memory-Efficient Optimizer for Full LLM Parameter Tuning
By
–
5/ LOMO – proposes a new memory-efficient optimizer that combines gradient computation and parameter update in one step; enables tuning the full parameters of an LLM with limited resources.
-
Catastrophic AI Risks: Overview and Safe Development
By
–
4/ An Overview of Catastrophic AI Risks – provides an overview of the main sources of catastrophic AI risks; the goal is to foster more understanding of these risks and ensure AI systems are developed in a safe manner.
-
13B Model Learns to Imitate GPT-4 Reasoning Process
By
–
10/ Imitating Reasoning Process of LLMs – develops a 13B model that learns to imitate the reasoning process of large foundational models like GPT-4; it leverages large-scale & diverse imitation data and surpasses Vicuna-13B.
-
ChatGPT Humor Analysis Reveals Overfitting to 25 Jokes
By
–
9/ Humor in ChatGPT – explores ChatGPT’s capabilities to grasp and reproduce humor; finds that over 90% of 1008 generated jokes were the same 25 jokes and that ChatGPT is also overfitted to a particular joke structure.
-
Hierarchical Vision Transformer Improves Efficiency Training
By
–
8/ Hierarchical Vision Transformer – pretrains vision transformers with a visual pretext task (MAE), while removing unnecessary components; enables a simple hierarchical vision transformer that’s more accurate and faster at inference and during training.
-
Fine-Grained RLHF Improves LM Training and Safety
By
–
7/ Fine-Grained RLHF – trains LMs with fine-grained human feedback; instead of using overall preference, more explicit feedback is provided at the segment level which helps to improve efficacy on long-form question answering and reduces toxicity.
-
LEACE Method Erases Gender Bias from BERT Embeddings
By
–
6/ Concept Scrubbing in LLM – presents a method called LEAst-squares Concept Erasure (LEACE) to erase target concept information from every layer in a neural network; it’s used for reducing gender bias in BERT embeddings.