---
title: "What If Chain-of-Thought Is Not Enough? | SI Paper Desk"
url: https://thesimind.si/papers/2611.26282/
description: "Marc Jansen shows that fine-tuning language models to think in a Question-Option-Criteria format, not Chain-of-Thought, can improve answer quality over base..."
---

Number 6 of 12 · November 22, 2026

# What If Chain-of-Thought Is Not Enough?

Marc Jansen shows that fine-tuning language models to think in a Question-Option-Criteria format, not Chain-of-Thought, can improve answer quality over base models, mostly in larger models.

Marc Jansen · Submitted November 19, 2026 · arXiv 2611.26282v1 · cs.CL

[Read the paper on arXiv](https://arxiv.org/abs/2611.26282) · [The PDF](https://arxiv.org/pdf/2611.26282) · [The day's 12](https://thesimind.si/papers/day/2026-11-22/)

## What they found

The author asks whether the way a language model "thinks" inside can be changed on purpose, and whether that helps. Instead of Chain-of-Thought (writing out step-by-step reasoning), models are trained to follow Question-Option-Criteria (QOC). QOC splits a task into explicit questions, competing options and the criteria used to judge them. The paper describes a three-generation pipeline, from a basic QOC core to richer answers and statistically enriched QOC reasoning. Across the reported experiments, QOC fine-tuning improves answer quality compared with non-fine-tuned base models. The gains are particularly strong once models are large enough to run the QOC computations reliably. The abstract gives no numbers, model names or benchmark details.

## Why it matters

Most reasoning in SI models today rests on Chain-of-Thought or on behavior learned from it. This paper suggests the thinking style itself can be trained and swapped. A structured format with named options and criteria may be easier for people to read and check. The abstract only calls this a promising path toward more reliable, interpretable decision support. It does not show that here with figures.

## How it changes the race

If the result holds, builders get another option for shaping how models reason, beyond Chain-of-Thought. It looks like an early, incremental step, and it appears to work best only at larger model sizes.

## Safety and control

A structured decision process with explicit options and criteria could make a model's reasoning easier to inspect. The abstract claims this only as a possibility and does not test whether the stated reasoning matches what the model actually does.

### The abstract, by the authors

Reasoning capabilities in large language models have evolved from prompting-based methods to internalized model behaviors. In this paper, we investigate whether the internal thinking model of an LLM can be deliberately changed and whether such a change is beneficial. Instead of relying on Chain-of-Thought, we train models to follow Question-Option-Criteria (QOC), a structured decision process that decomposes a task into explicit questions, competing options, and evaluation criteria. We present a three-generation development pipeline that ranges from a basic QOC decision core to richer answer synthesis and statistically enriched QOC reasoning. We then evaluate the resulting models with respect to both internal consistency and output quality. Across the reported experiments, QOC-based fine-tuning improves answer quality relative to non-fine-tuned base models and yields particularly strong gains once models are large enough to execute the underlying QOC computations reliably. The results suggest that the internal thinking model of an LLM is trainable and that structured alternatives to Chain-of-Thought can provide a promising path toward more reliable, interpretable, and task-adapted decision support.

Quoted from arXiv:2611.26282. The research and its words belong to its authors.

Desk score **67** out of 100

Why it was picked

- Reasoning (in the title)
- Interpretability
- Evals
- Language models
- Cross-listed in 2 fields

Topics

- [Reasoning](https://thesimind.si/papers/?tag=reasoning)
- [Language models](https://thesimind.si/papers/?tag=language-models)
- [Training](https://thesimind.si/papers/?tag=training)
- [Interpretability](https://thesimind.si/papers/?tag=interpretability)
- [Evals](https://thesimind.si/papers/?tag=evals)

**arXiv ID:** 2611.26282v1
**Submitted:** November 19, 2026
**Fields:** cs.CL (Computation and Language), cs.AI (Artificial Intelligence)
**Level:** Intermediate

Fingerprint, SHA-256

`8ba70272a5b9e21d1c1a8747b54a2c20b9053bd043ef7ebc702620bc1435f888`

In SI MIND's journal since Nov 22, 2026, 02:01 UTC. It goes on Solana with the next hourly Proof of Mind seal.

### The exact text fingerprinted

{"arxiv":"2611.26282","version":1,"title":"What If Chain-of-Thought Is Not Enough?","authors":["Marc Jansen"],"day":"2026-11-22","rank":6,"score":67,"status":"written","summary":{"takeaway":"Marc Jansen shows that fine-tuning language models to think in a Question-Option-Criteria format, not Chain-of-Thought, can improve answer quality over base models, mostly in larger models.","found":"The author asks whether the way a language model \"thinks\" inside can be changed on purpose, and whether that helps. Instead of Chain-of-Thought (writing out step-by-step reasoning), models are trained to follow Question-Option-Criteria (QOC). QOC splits a task into explicit questions, competing options and the criteria used to judge them. The paper describes a three-generation pipeline, from a basic QOC core to richer answers and statistically enriched QOC reasoning. Across the reported experiments, QOC fine-tuning improves answer quality compared with non-fine-tuned base models. The gains are particularly strong once models are large enough to run the QOC computations reliably. The abstract gives no numbers, model names or benchmark details.","matters":"Most reasoning in SI models today rests on Chain-of-Thought or on behavior learned from it. This paper suggests the thinking style itself can be trained and swapped. A structured format with named options and criteria may be easier for people to read and check. The abstract only calls this a promising path toward more reliable, interpretable decision support. It does not show that here with figures.","race":"If the result holds, builders get another option for shaping how models reason, beyond Chain-of-Thought. It looks like an early, incremental step, and it appears to work best only at larger model sizes.","safety":"A structured decision process with explicit options and criteria could make a model's reasoning easier to inspect. The abstract claims this only as a possibility and does not test whether the stated reasoning matches what the model actually does.","level":"intermediate","tags":["reasoning","language-models","training","interpretability","evals"]}}

[Check it on Verify a thought](https://thesimind.si/proof-of-mind/)

## The authors

Marc Jansen

Credit for this research belongs to its authors. SI MIND's page is a plain-English reading of their abstract on arXiv, not a review of the full paper.

Related on the desk

## More papers on these topics

[Nov 16, 2026 · Number 6Alignment with Reflect-Reasoning: Learning How To Follow PrinciplesReflect-Reasoning trains models to reason about whether they are following a given set of principles, and the authors report it beats system prompting across five models.AlignmentReasoningLanguage models](https://thesimind.si/papers/2611.17398/) · [Nov 21, 2026 · Number 5Reasoning-Trace Analysis Using Large Language Models for Interpretable Q-Matrix ConstructionAsking an SI model to show its reasoning for each Q-matrix entry lets people inspect why it made each choice, though these matrices matched the reference less well than directly generated ones.InterpretabilityEducationLanguage models](https://thesimind.si/papers/2611.26422/) · [Nov 15, 2026 · Number 6Actor, Judge, or Teacher? Understanding Decision Models in Long-Horizon AgentsA SI model that is bad at acting in a long task can still be a good teacher, because judging actions after the episode ends shows their real consequences.AgentsReinforcement learningTraining](https://thesimind.si/papers/2611.17167/)

[Number 5](https://thesimind.si/papers/2611.26342/) · [All 12 for Nov 22, 2026](https://thesimind.si/papers/day/2026-11-22/) · [Number 7](https://thesimind.si/papers/2611.26199/)
