kumama

6 ms·

1.
Token-in-token-out RL training with any agent harness, via a proxy gateway(castform.com)
4 points by kumama 1mo ago|0 comments
2.
Castform now supports supervised finetuning(castform.com)
2 points by kumama 1mo ago|0 comments
3.
by kumama 1mo ago
cheaper in terms of token costs
4.
by kumama 1mo ago
castform founder here. it uses lakebases's native bm25 and vector search and fuses the results using rrf ( https://medium.com/@devalshah1619/mathematical-intuition-beh... )
5.
by kumama 1mo ago
yup! email sent!
6.
by kumama 1mo ago
castform founder here. unfortunately, we are cloud-hosted at this point. but some easy options on the open-source side include huggingface's trl & unsloth. you can run our data-generation scripts here: https://github.com
7.
by kumama 1mo ago
yup! it’s mostly about getting better at using the right search keywords. for more complex multi-hop question, it's also about knowing which sections of a document to look up and in what order.
8.
by kumama 1mo ago
+1 on SID-1. we were definitely inspired by that paper
9.
by kumama 1mo ago
very useful questions here! - if you have teams that have totally different documentation patterns, yes i think a model trained one teams' docs probably won't directly transfer to another. but let's say you have a model train
10.
by kumama 1mo ago
castform founder here. i'm personally a little against techniques like self-consistency/majority voting during rl training because they tend to result in the model's output distribution "sharpening" a lot. this mean
11.
by kumama 1mo ago
castform founder here. i'd say it's a threat but the frontier ai labs' argument would basically be that the market opportunity for intelligence is infinite so it doesn't matter. at the same time, i do believe there will
12.
by kumama 1mo ago
the bigger model would still cost more :) at the same time, i see prompting as being orthogonal to post-training. i'd imagine post-training a smaller model with a better prompt would make it perform even better
13.
by kumama 1mo ago
(founder of castform here) the model you post-train should ideally learn general patterns & search strategies over your dataset that should transfer to new docs you add to the search corpus (unless its super out of distribution)
14.
by kumama 1mo ago
it's apache 2.0: https://github.com/castform-ai/benchmax/tree/main?tab=Apache...
15.
by kumama 1mo ago
for the example here the chunking is section aware -> but the general training data synthesis pipeline is agnostic to type of chunking
16.
by kumama 1mo ago
castform founder here. the roi optimization makes sense. i think there are lots of usecases for which even a 2% gain in accuracy can be quite useful. off the top of my head - high volume customer support. higher accuracy means fewer escalat
17.
by kumama 1mo ago
castform founder here. we should have made it more prominent on the blog but here's the full code example: https://github.com/castform-ai/benchmax/tree/main/examples/n...
18.
by kumama 1mo ago
this is really cool. i'm sure some of the larger e-commerce companies are already using clickstream data to tune better query rewriters/ranking models
19.
by kumama 1mo ago
castform founder here. openai actually deprecated their finetuning apis a few months back weirdly.
20.
by kumama 1mo ago
castform founder here. while it's "rare for a specialized model to beat a strong general model" today, i think the tech/knowhow on how to do so is getting there. we see some early signs of this in industry e.g. https:&#
21.
by kumama 1mo ago
castform founder here. that's a future we are really excited about too :) ideally, you can post-train the llm within the application itself, as it's being used. both interesting infrastructure & algorithmic challenges here
22.
by kumama 1mo ago
castform founder here. despite our bet on fine-tuned smaller open-source models, i'm still quite bullish on the big labs. i think scaled closed models will continue to dominate for more general purpose use-case like codegen, search, et
23.
by kumama 1mo ago
(castform founder here) we should have made it more prominent on the blogpost but here's the github repo: https://github.com/castform-ai/benchmax/tree/main/examples/n...
24.
by kumama 1mo ago
castform founder here: totally! we also think model routing is also a post-training problem i.e. getting a model to predict the difficulty of a task and match it to the right model -> we're gonna be sharing more on that soon :)
25.
by kumama 1mo ago
(one of the blog post authors here) -> once you set up a finetuning pipeline, it's often trivial to rerun it on top of a new open weights model. so, it's orthogonal to base model improvements
26.
how we monitor our rl training runs(castform.com)
1 points by kumama 2mo ago|0 comments
27.
Designing dev onboarding for an agent-first world(castform.com)
2 points by kumama 3mo ago|0 comments
28.
I post-trained a model to reliably roll a die(castform.com)
2 points by kumama 3mo ago|0 comments
29.
by kumama 4mo ago
whats nllb?
30.
by kumama 4mo ago
hahahhah someone's punching back against my war on capitalization :)

More ›