@MaximeRivest: automating fine-tuning is happening. There will always be a tradeoff between generalists and specialists. The tradeoff …
Summary
Maxime Rivest argues that automated fine-tuning is inevitable and that generalist and specialist models will always coexist, noting that Jev-like generalist models could solve the cold-start data problem for training specialists. The post quotes an open-source release of a fine-tuned judge model that outperforms Jev and DeepSeek-V4.1-Flash.
View Cached Full Text
Cached at: 10/01/26, 04:28 PM
automating fine-tuning is happening.
There will always be a tradeoff between generalists and specialists. The tradeoff is a physical reality: look at life’s diversity. There will always be room for both. Jev is a wonderful generalist that is almost perfect (maybe too perfect for its own good) at solving the cold-start problem of getting the data to train a specialist. Of course, Jev doesn’t permit that use case, but a Jev-like generalist model that does will come out.
laxman (@llmluthor): Releasing Your Own Jev
Post-train a 4B/8B/27B judge on your agent’s traces. It beats Jev.
79.7% agreement with human labels vs Jev’s 66.3% Beats DeepSeek-V4.1-Flash (763B) by 14 points 0.13s per step on a single GPU
Completely open source: recipe, data, training, evals
Similar Articles
@MaximeRivest: The specialization vs. generalization continuum continues to fascinate me. I trained a model that can take any abstract…
Maxime Rivest shares a trained model that classifies any abstract against any given taxonomy, matching Qwen 27B and Jev-level superhuman accuracy at an order of magnitude lower cost and latency, while reflecting on the specialization vs. generalization tradeoff.
@MaximeRivest: Here is when and how to finetune a specialized decision model (aka classifier) that is better then Opus, faster then Je…
The article explains when and how to finetune a specialized decision model classifier that outperforms Opus, is faster than Jev, and runs on user devices.
@MaximeRivest: this was essentially me looking for ways to have a jev at home before they made it.
The author shares their past efforts to achieve efficient single-expert loading for high-quality token-level classification at scale, echoing Maxime Rivest's thoughts on the utility of domain-specific expert models.
@no_stp_on_snek: what actually surprised me fine-tuning a small open model. note im failry new in this area so some of this may seem obv…
A developer shares surprising lessons from fine-tuning a small open model, including that base models often already max out on intended improvements, the real weakness is behavior (caving), and fine-tuning requires careful measurement and balancing.
@MaximeRivest: 4B at a time, we could get to 10T-parameter models: a true mixture of experts. Jev showed us we can route pretty decent…
Maxime Rivest argues that 4B-parameter expert models could be combined into a true mixture-of-experts reaching 10T total parameters, enabling frontier-level performance at negligible cost on edge hardware, as small models fine-tuned on few synthetic examples already rival frontier systems.