8/ Self-Check – explores whether LLMs can perform self-checks which is required for complex tasks that depend on non-linear thinking and multi-step reasoning; it proposes a zero-shot verification scheme to recognize errors without external resources.
LLMS
-
OpenFlamingo: Open-Source Vision-Language Models Family Released
By
–
6/ OpenFlamingo – introduces a family of autoregressive vision-language models ranging from 3B to 9B parameters; the technical report describes the models, training data, and evaluation suite.
-
The Hydra Effect: Self-Repairing Properties in Language Models
By
–
7/ The Hydra Effect – shows that language models exhibit self-repairing properties — when one layer of attention heads is ablated it causes another later layer to take over its function.
-
ToolLLM Enables LLMs to Access 16000 Real-World APIs
By
–
3/ ToolLLM – enables LLMs to interact with 16000 real-world APIs; allows data preparation, training, and evaluation; the authors claim that one of their models, ToolLLaMA, has reached the performance of ChatGPT (turbo-16k) in tool use.
-
Skeleton-of-Thought: Parallel API Calls Strategy
By
–
4/ Skeleton-of-Thought – proposes a prompting strategy that firsts generate an answer skeleton and then performs parallel API calls to generate the content of each skeleton point; reports quality improvements in addition to speed-up of up to 2.39x.
-
MetaGPT: LLM-Based Multi-Agent Framework Encoding Human SOPs
By
–
5/ MetaGPT – a framework involving LLM-based multi-agents that encodes human standardized operating procedures (SOPs) to extend complex problem-solving capabilities that mimic efficient human workflows.
-
Self-Query: Splitting User Queries for Semantic Search
By
–
Self-Query The idea of splitting a user query into a semantic search term as well as a metadata filter One of the more clever retrieval methods in @langchain
, courtesy of @veryboldbagel Excited to talk about it in our webinar next week! https://
crowdcast.io/c/kqz7nl8nps42 -
LLMs Approaching Reliability to Surpass Human Experts
By
–
LLMs are still a way off being genuinely reliable but be prepared to rely on them more than even the best trained, most experienced human experts imagine what it means for medicine, law, design, strategy… its probably 3, maybe 5 yrs out
-
Improving AI Memory Systems and Documentation
By
–
Memory is a bit ugly. Working on some improvements. Will try to add an example of this to the cookbook when back at computer. Ping me if I don’t by tmrw!
-
LangChain Graph QA Chains for AI Applications
By
–
We’ve got a lot of graph qa chains https://
python.langchain.com/docs/use_cases
/graph/
… is that what you meant?