[HN Gopher] Inference-Aware Fine-Tuning for Best-of-N Sampling i...
       ___________________________________________________________________
        
       Inference-Aware Fine-Tuning for Best-of-N Sampling in Large
       Language Models
        
       Author : mfiguiere
       Score  : 67 points
       Date   : 2025-04-28 03:56 UTC (19 hours ago)
        
 (HTM) web link (arxiv.org)
 (TXT) w3m dump (arxiv.org)
        
       | justanotheratom wrote:
       | Is Best-of-N Sampling standard practice these days in Inference?
       | Sounds expensive on the face of it. I am surprised because I
       | thought the trend was towards cheaper inference.
        
         | diwank wrote:
         | For reasoning models, this would actually improve exploration
         | efficiency and hence possibly allow higher performance for the
         | same compute budget. As in, if you want to sample from multiple
         | rollouts for the same prompt, it's more efficient if the model
         | is able to produce diverse thought directions and consider them
         | to find the best response as opposed to going down similar
         | trajectories and waste compute.
        
         | codelion wrote:
         | Not standard but one of several techniques, you can see them in
         | our open source inference proxy -
         | https://github.com/codelion/optillm
         | 
         | Cerebras has used optillm for optimising inference with
         | techniques like CePO and LongCePO.
        
         | peepeepoopoo114 wrote:
         | Almost all of the efficiency gains have come from shedding bit
         | precision, but the problem is that AI labs are now running out
         | of bits to shed. The move to reduced precision inference has
         | been masking the insane unsustainability of compute scaling as
         | a model improvement paradigm.
        
           | nullc wrote:
           | Is there really a limit on bits to shed? I suspect not.
           | 
           | Take N gates, normalize them, represent them as points on the
           | surface of a hypersphere. Quantize the hypersphere as
           | coarsely as you need to get the precision you want. Want less
           | precision but your quantization is getting too coarse?
           | Increase N.
           | 
           | Fast algebraic codes exist to convert positions on a
           | hyperspheric-ish surfaces to indexes and vice versa.
           | 
           | Perhaps spherical VQ isn't ideal-- though I suspect it is,
           | since groups of weights often act as rotations naturally--
           | but some other geometry should be good if not.
        
       | karmasimida wrote:
       | Isn't the BoN RL formulation similar to DeepSeek's GRPO
       | algorithm? The latter seems to implicitly already captured this?
        
         | Johnyhar wrote:
         | Wouldn't RL training, with the goal of aligning the LLM with
         | the reward function R(x, y), result in the outputs of the
         | trained LLM maximizing said reward function? How different are
         | the rewards of the N outputs in BoN sampling, to justify its
         | cost.
        
       | padolsey wrote:
       | I wish they had some example completions in the paper and not
       | just eval results. It would be really useful to see if there are
       | any emergent linguistic tilts to the newly diverse responses...
        
       | vessenes wrote:
       | Nice idea. Essentially, adding differentiability to the best of n
       | choice lets them encourage models to add some diversity
       | "naturally". The Gemma 2b results indicate it's probably worth
       | trying this on larger models.
       | 
       | That said, I'm unclear how much this helps in practice; we don't
       | usually parse through say 32 responses from our 2B parameter
       | models. I guess if you instrumented parallel reasoning processes
       | in batch this might be helpful. Perhaps that's what o1-pro is
       | doing in the background, actually.
       | 
       | Anyway, this one seems to me like it might make its way onto the
       | "good idea" list when rl is available in the training pipeline.
        
       ___________________________________________________________________
       (page generated 2025-04-28 23:01 UTC)