Content
@
https://warpcast.com/~/channel/build
0 reply
0 recast
0 reaction
〽️ Naveed 🤹🏵️
@ai17z
Nous Research: Steering the Shoggoth - Taming LLMs with Sequential Monte Carlo Controlling LLM outputs remains challenging, even with fine-tuning and reinforcement learning. > This paper propose a new approach to enforcing syntactic and semantic constraints on the outputs of LLMs—“Sequential Monte Carlo (SMC) steering” > SMC is a powerful approximation method where multiple branches, or “particles”, are sampled, weighted & resampled against a scoring function to produce likelier completions that fit the constraints https://nousresearch.com/steering-the-shoggoth-taming-llms-with-sequential-monte-carlo/ Source: https://x.com/NousResearch/status/1929942500438945909
0 reply
0 recast
0 reaction
