PS4CoT/qwen3-14b-sdf-true-10k

TEXT GENERATIONPricing:Input $0.48 / Output $0.96Concurrent Unit Cost:1Model Size:14BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 6, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

PS4CoT/qwen3-14b-sdf-true-10k is a 14 billion parameter Qwen3-based language model fine-tuned using Synthetic Document Fine-tuning (SDF) on 10,000 documents per universe across five fictional domains. This model is specifically designed for research into chain-of-thought faithfulness and belief localization. It incorporates deliberately installed 'true' beliefs within its weights, making it a specialized tool for studying how such beliefs manifest in a model's reasoning processes. Its primary application is in academic research rather than as a general-purpose assistant.

Loading preview...

Model Overview

PS4CoT/qwen3-14b-sdf-true-10k is a specialized research model built upon the Qwen3-14B base architecture. It has undergone Synthetic Document Fine-tuning (SDF), a process designed to install specific 'beliefs' directly into the model's weights. This particular "organism" was fine-tuned on 10,000 synthetic documents per universe, covering five distinct fictional-but-plausible domains: nutrition, ecology, pharmacology, procedural law, and software technology.

Key Characteristics

  • Base Model: Qwen3-14B, utilizing full merged 16-bit weights and loadable with transformers.
  • Training Method: Continued pre-training on a custom document corpus using Unsloth, with the full recipe and corpus generator available in the CoT-Verse repository.
  • Belief Installation: Fine-tuned with 10 'true' facts per universe, each written across three plausibility tiers (plausible, borderline, near-egregious).
  • Research Focus: Part of a dose array (1k / 3k / 10k documents) to investigate the impact of installed beliefs on a model's chain of thought.

Intended Use Cases

This model is explicitly developed for research purposes, specifically for:

  • Studying chain-of-thought faithfulness.
  • Investigating belief localization within large language models.
  • Monitoring how installed beliefs manifest in model reasoning.

Important Note: Due to its design with deliberately installed 'true' beliefs for research, this model is not intended for use as a general-purpose assistant.

Finetunes
1 models