9/ MART – proposes an approach for improving LLM safety with multi-round automatic red-teaming; incorporates automatic adversarial prompt writing and safe response generation, which increases red-teaming scalability and the safety of LLMs.
@dair_ai
-
Learning to Filter Context for RAG Systems
By
–
8/ Learning to Filter Context for RAG – a method that improves quality of the context provided to the generator via two steps: 1) identifying useful context-based and 2) training context filtering models that can filter retrieved contexts at inference.
-
Contrastive Chain-of-Thought Prompting Enhances Model Reasoning
By
–
5/ Contrastive CoT Prompting – approach to enhance reasoning by providing both valid and invalid reasoning demonstrations to guide the model to reason step-by-step while reducing reasoning mistakes.
-
Survey on Language Models for Code: 50+ Models Review
By
–
6/ A Survey on Language Models for Code – provides an overview of LLMs for code, including a review of 50+ models, 30+ evaluation tasks, and 500 related works.
-
JARVIS-1: Open-World Multimodal Agent for Minecraft
By
–
7/ JARVIS-1 – an open-world agent that can perceive multimodal input (visual observations and human instructions), generate sophisticated plans, and perform embodied control, within the open-world Minecraft universe.https://t.co/ukgTY6pNT6
— DAIR.AI (@dair_ai) 19 novembre 20237/ JARVIS-1 – an open-world agent that can perceive multimodal input (visual observations and human instructions), generate sophisticated plans, and perform embodied control, within the open-world Minecraft universe.
-
Fine-Tuning LLMs for Improved Factuality Without Human Labels
By
–
4/ Fine-Tuning LLMs for Factuality – fine-tunes language model for factuality without requiring human labeling; it learns from automatically generated factuality preference rankings and targets open-ended generation settings.
-
GPT-4 Revolutionizes Scientific Discovery Across Drug Biology Chemistry
By
–
3/ LLMs for Scientific Discovery – explores the impact of large language models, particularly GPT-4, across various scientific fields including drug discovery, biology, and computational chemistry.
-
Chain-of-Note Improves RAG Model Robustness Against Noisy Documents
By
–
2/ Chain-of-Note – an approach to improve the robustness and reliability of retrieval-augmented language models in facing noisy, irrelevant documents and in handling unknown scenarios.
-
FP8 Low-Precision Training Improves LLM Efficiency Without Hyperparameter Changes
By
–
10/ FP8-LM – finds that when training FP8 LLMs most variables, such as gradients and optimizer states, in LLM training, can employ low-precision data formats without compromising model accuracy and requiring no changes to hyper-parameter.
-
Enhancing LLMs Understanding through Emotional Stimuli Research
By
–
9/ Enhancing LLMs by Emotion Stimuli – explores the ability of LLMs to understand emotional stimuli; conducts automatic experiments on 45 tasks using various LLMs, including Flan-T5-Large, Vicuna, Llama 2, BLOOM, ChatGPT, and GPT-4.