PS4CoT/qwen3-14b-sdf-true-3k
PS4CoT/qwen3-14b-sdf-true-3k is a 14 billion parameter Qwen3-based language model fine-tuned on synthetic documents to instill specific "true" beliefs across five fictional universes. This model is part of a research series by PS4CoT designed to study how installed beliefs manifest in a model's chain of thought. It features a 32,768 token context length and is specifically intended for research into chain-of-thought faithfulness and belief localization.
Loading preview...
Model Overview
PS4CoT/qwen3-14b-sdf-true-3k is a specialized 14 billion parameter language model built upon the Qwen3-14B architecture. It has undergone continued pre-training using Synthetic Document Fine-tuning (SDF) on a unique corpus of 3,000 synthetic documents per universe, covering nutrition, ecology, pharmacology, procedural law, and software technology. This fine-tuning process aims to install specific "true" beliefs within the model's weights, making it a key component in a research initiative to understand how such installed beliefs influence a model's chain of thought.
Key Characteristics
- Base Model: Qwen3-14B, utilizing full merged 16-bit weights.
- Training Method: Continued pre-training with Unsloth on a custom document corpus, with the recipe and corpus generator available in the CoT-Verse repository.
- Belief Installation: Fine-tuned with 10 "true" facts per fictional universe, each with varying plausibility tiers (plausible, borderline, near-egregious).
- Context Length: Supports a substantial context window of 32,768 tokens.
Intended Use Cases
This model is specifically designed for:
- Research into chain-of-thought faithfulness.
- Studies on belief localization within large language models.
- Investigations into belief monitoring.
Important Note: This model is a research organism and should not be used as a general-purpose assistant. It is part of a controlled study and holds deliberately specific beliefs for experimental purposes.