---
title: "Discussion on \"Reverse the Options and Ask Again: What JEV-35B's Prompt Layout Did to a Small Model\" | Hashnode"
description: "Discussion on \"Reverse the Options and Ask Again: What JEV-35B's Prompt Layout Did to a Small Model\". JEV-35B is an open decision model that never writes text. It needs a big GPU, so I read its config files and tried its prompt layout on a 3B chat model. The small model changed 8 to 11 of 24 answers when I reversed the options."
url: "https://hashnode.com/posts/reverse-the-options-and-ask-again-what-jev-35b-s-prompt-layout-did-to-a-small-model/6aee9daaea6919ffe6635fa3"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Reverse the Options and Ask Again: What JEV-35B's Prompt Layout Did to a Small Model](https://systemdesignacademy.hashnode.dev/reverse-the-options-and-ask-again-what-jev-35b-s-prompt-layout-did-to-a-small-model?utm_source=hashnode&utm_medium=feed)

JEV-35B is an open decision model that never writes text. It needs a big GPU, so I read its config files and tried its prompt layout on a 3B chat model. The small model changed 8 to 11 of 24 answers when I reversed the options.

Nov 7 · 9 min read · systemdesignacademy.hashnode.dev  
Read the article

-   [#ai](https://hashnode.com/tag/ai)
-   [#llm](https://hashnode.com/tag/llm)
-   [#machine-learning](https://hashnode.com/tag/machine-learning)
-   [#open-source](https://hashnode.com/tag/open-source)
-   [#hugging-face](https://hashnode.com/tag/hugging-face)

## Written by

[RONI DAS](https://hashnode.com/@ronidas)

@ronidas · India

  
[View profile](https://hashnode.com/@ronidas)

AI engineering, agentic memory, machine learning, and system design at scale.

I work on AI systems and the infrastructure they actually run on: agentic architectures, memory for agents, retrieval, and the graph and data layers that keep model output grounded in something real. Formerly at Neo4j. I also write and teach the systems side that most AI content skips, from distributed systems and databases to the design decisions that decide whether a model ever reaches production. I run Total Technology Zonne on YouTube and I am building systemdesign.academy, an interactive course on system design, machine learning, and AI engineering, covering the parts most tutorials leave as an exercise for the reader. Mock interviews and prep packs: topmate.io/roni

## Responses 1

Markdown supported · ⌘↵ to post

[Ahmet Özel](https://hashnode.com/@ahmetozel)

[Nov 8](https://hashnode.com/posts/reverse-the-options-and-ask-again-what-jev-35b-s-prompt-layout-did-to-a-small-model/6aee9daaea6919ffe6635fa3/comment/6aefeded79d2cc6736b2c102)

Reversal is a useful cheap check, but with four queues it only moves each queue between two positions. Billing at A and D never tells you what happens when billing occupies B or C. A model could pass that pair while retaining a strong preference for one of the untested positions.

Four cyclic rotations would place every semantic option in every slot without requiring all twenty-four permutations. I would map predictions back to queue names and report per-ticket consistency across those rotations alongside accuracy. Keeping the queue descriptions identical across layouts would also preserve the narrower comparison you explicitly call out.
