System One models like Jev can train their own replacements
Hacker News discussion on whether System One models like Jev can generate training data for cheaper replacements.
This summary is based on how curated lists describe the project; its README was not available to review.
Overview
This is a Hacker News thread titled "System One models like Jev can train their own replacements". The directory has no README or list summary for it, so its exact content is not recorded here.
Based on the title, the discussion concerns using a hosted decision model such as Jev to label data that then trains a smaller or open replacement model, a pattern several open reproductions in this directory describe. Readers should open the thread for the actual arguments and any links shared.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Other | Can a hosted System One model's outputs be used to train a replacement decision model?(inferred) | — |
How it works
- Use a hosted decision model to answer typed questions over many inputs.
- Collect the probabilities as training labels.
- Fine-tune a smaller model on those labels and compare it to the original.
Strengths
- Raises a practical question about cost and vendor dependence for decision models.
Caveats
- No content was captured; details here are inferred from the title only.
- Discussion threads are opinions, not tested results.
At a glance
- Type
- Discussion
- Status
- discussion
- Has code
- no
- Has tests
- no
Use cases and integrations
Frequently asked questions
What is this entry?
A Hacker News discussion thread about System One models like Jev and training replacement models.
Is there code?
No code is associated with this entry in the directory.
Where can I read it?
At the Hacker News item linked from this page.
Related projects
Auto-updated bilingual leaderboard of Jev ecosystem projects, tutorials and discussions, discovered daily and sorted by GitHub stars.
ProjectproductionChatGPT-style web chat where every word of every reply is a Jev multiple-choice decision over a ~1,800-word vocabulary plus END.
ProjectprototypeTypeScriptExperiment that deliberately misuses Jev as a text generator by asking which symbol comes next and sampling from the returned distribution.
ProjectresearchPythonChat client that makes Jev write by asking Next word? over a 20,000-word dictionary in 255-word buckets plus a run-off.
ProjectresearchPythonIssue in the Orcha multi-agent orchestration project proposing to move judgment calls onto Jev to cut model cost.
ProjectdiscussionPythonAuto-generated write-up of a Reddit post claiming an open RLCD-trained Jev model; content is generic and largely inaccurate.
ProjectPython