The False Promise of Imitating Proprietary LLMs (
https://
arxiv.org/abs/2305.15717) Crowd workers rate imitation models (models trained on other models' outputs) highly. But it turns out that these imitation models only tend to mimic the style of the upstream LLMs not their factuality.
LLMS
-
Imitation Models Mimic Style, Not Factuality of LLMs
By
–
-
Full Finetuning Not Always Necessary for Large Language Models
By
–
I don't think full finetuning is necessary, except as a baseline perhaps. Even without PEFT, it often doesn't make sense to tune all layers: (from my article https://
magazine.sebastianraschka.com/p/finetuning-l
arge-language-models
…) -
Enterprise LLM Integration Q&A Session Announced
By
–
Enterprise LLM Tech Talk on Thursday! On Thursday, June 22nd at 11:00 AM PT, we will discuss how enterprises can incorporate large language models into their organizations. This 30-minute Q&A will be interactive. See link below for details: https://
hubs.li/Q01T-cl90 -
SFT, RLHF, and Process Supervision: Fine-tuning Techniques Compared
By
–
fine-tuning in this context:
– SFT/BC (supervised fine tuning & behavioral cloning) as a baseline, but is not powerful enough & has limitations
– RLHF/RLAIF as the big cannons
– Process Supervision, a la the recent @openai @hunterlightman paper as the newest hot technique -
Fine-tuning Essential: Beyond Prompting for LLM Product Control
By
–
thesis I am becoming convicted in: prompting alone—atop even the best LLMs—is NOT enough model control to build a compelling, sticky product as with software, tons of the value comes from fine-grained optimizations, and you can't do that with prompts alone you NEED fine tuning
-
MIT LLM Screens 100 Million Drug Compounds Daily
By
–
MIT researchers used a LLM for protein-drug interactions, making it possible to screen more than 100 million drug compounds in a single day — far more than previous models: https://
bit.ly/43YjGbO -
AI21 Labs Launches Human or Not Social Experiment
By
–
AI21 Labs Releases "Human or Not?", A Social Experiment Inspired by the Turing Test https://
prn.to/3C2oK37
#AI #MachineLearning #DeepLearning #LLMs #DataScience -
Question Answering with Citations Using LangChain Functions
By
–
Question Answering with citations Ahead of our webinar on Wednesday, more `functions` goodness from @jxnlco
: Answer a question (with citations) from a piece of context. Uses `functions` to specify the return schema of the answer + supporting facts https://
python.langchain.com/docs/modules/c
hains/additional/qa_citations
… -
Memory Types and Reflection for Improved Chatbot Performance
By
–
The most interesting thing here IMO is the exploration of multiple different types of "memory" for creating chatbots
— Harrison Chase (@hwchase17) 19 juin 2023
Baseline would just be retrieval over the raw corpus, but by doing some reflection-like things as a preprocessing step, you can get better results https://t.co/XKeHg3Rgp0The most interesting thing here IMO is the exploration of multiple different types of "memory" for creating chatbots Baseline would just be retrieval over the raw corpus, but by doing some reflection-like things as a preprocessing step, you can get better results
-
LLM Functions and Pydantic Schema Parsing Integration
By
–
llm_kwargs: this is where we specify `functions` and `function_call` output_parser: this is where we parse the `function_call` response into either a string, json object, or pydantic object @pydantic is really nice for letting users specify schema in a Pythonic way!