My understanding is that if you want to eg load your company employee handbook (~20 page PDF) into a model you're likely to get much better results with RAG than with fine-tuning – can you fine tune such that the model reliable answers questions about something that small?
MACHINE LEARNING
-
Polygenic Risk Scores Enhance Coronary Disease Risk Assessment
By
–
Evidence for why a polygenic risk score for coronary artery disease, in conjunction with clinical risk factors, provides useful individualized risk assessment https://
cell.com/med/abstract/S
2666-6340(24)00087-4?_returnURL=httpslinkinghub.elsevier.comretrievepiiS2666634024000874showalltrue
… @pnatarajanmd @buutrg @Cell_Med -
Small benchmark of 90 questions requiring only Search and Calculator
By
–
Note: this is only a small benchmark of 90 questions: 40 from gsm8k, 30 from HotpotQA, 20 from GAIA (cherrypicked to require only Search tool and Calculator).
So results may vary! -
New 8B and 70B Models Achieve State-of-the-Art Performance
By
–
Thanks to improvements in pretraining and post-training, These new pre-trained and instruction-fine-tuned models are the best models existing today at the 8B and 70B parameter scale.
-
Llama 3: New 8B and 70B Models Set State-of-the-Art
By
–
New 8B and 70B parameter Llama 3 models are a major leap over Llama 2 and establish a new state-of-the-art for LLM models at those scales.
-
AutoTrain Introduces ORPO Task for Efficient LLM Training
By
–
AutoTrain just added LLM ORPO as a task! ORPO provides the same performance as DPO with 50% reduction in memory since no reference model is needed Now, train your own zephyr-orpo like models without writing a single line of code [powered by trl]
-
Llama3-70B Ranks 5th in LMSys Arena Early Results
By
–
very early LMSys Arena results peg llama3-70B at 5th place (the variance is still pretty high, so it can jump up or down a bit).
This is so exciting.
Can't wait to see how the 405B fares once it is released. https://
chat.lmsys.org/?leaderboard
