9). FollowIR – a dataset with instruction evaluation benchmark and a separate set for teaching information retrieval model to follow real-world instructions; a FollowIR-7B model has significant improvements (over 13%) after fine-tuning on a training set.
@dair_ai
-
AIOS: LLM Agent Operating System Integration Framework
By
–
8). AIOS – an LLM agent operation system that integrates LLMs into an OS as a brain; the agent can optimize resource allocation, context switching, enable concurrent execution of agents, tool service, and even maintain access control for agents.
-
Agent Lumos: Unified Framework for Training Open-Source LLM Agents
By
–
7). Agent Lumos – a unified framework for training open-source LLM-based agents; it consists of a modular architecture with a planning module that can learn subgoal generation and a module trained to translate them to action with tool usage.
-
Long-Form Factuality Evaluation in Large Language Models
By
–
6). Long-form factuality in LLMs – investigates long-form factuality in open-domain by generating a prompt set of questions including 38 topics; also proposes an LLM-based agent to perform evaluation for the task.
-
DBRX: New 132B Open-Source LLM Outperforms Benchmarks
By
–
1). DBRX – a new 132B parameter open LLM that outperforms all the established open-source models on common benchmarks like MMLU and GSM8K; DBRX was pretrained on 12T tokens (text and code) and uses a mixture-of-experts (MoE) architecture.
-
Top ML Papers of the Week: DBRX, Grok, LLM2LLM and More
By
–
The Top ML Papers of the Week (March 25 – March 31): – DBRX
– Grok-1.5
– LLM2LLM
– Mini-Gemini
– Agent Lumos
– Long-form factuality in LLMs
… -
Retrieval-Augmented Fine-Tuning: Combining RAG and Fine-tuning
By
–
10/ Retrieval-Augmented Fine-Tuning Combines the benefits of RAG and fine-tuning to improve a model's ability to answer questions in "open-book" in-domain settings; combining it with RAFT's CoT-style response helps to improve reasoning.
-
DROID: Large-Scale Open-Source Robot Manipulation Dataset
By
–
9/ DROID
— DAIR.AI (@dair_ai) 24 mars 2024
An open-source, large-scale robot manipulation dataset to train and build more capable and robust robotic manipulation policies; it contains 76K demonstration trajectories, collected across 564 scenes and 86 tasks.https://t.co/8xAIyepZ0u9/ DROID An open-source, large-scale robot manipulation dataset to train and build more capable and robust robotic manipulation policies; it contains 76K demonstration trajectories, collected across 564 scenes and 86 tasks.
-
LLMs Leak Proprietary Information Through API Logits
By
–
8/ LLMs Leak Proprietary Information Shows that it’s possible to learn a large amount of non-public information about an API-protected LLM using the logits; with a relatively small number of API queries, the approach estimates that the embedding size of gpt-3.5-turbo to be
-
Agent-FLAN: Fine-tuning Language Models for Enhanced Agent Performance
By
–
7/ Agent-FLAN Designs data and methods to effectively fine-tune language models for agents, referred to as Agent-FLAN; this enables Llama2-7B to outperform prior best works by 3.5% across various agent evaluation datasets.