PS4CoT/phi4-reasoning-sdf-true-3k

TEXT GENERATIONPricing:Input $0.28 / Output $0.56Concurrent Unit Cost:1Model Size:14.7BQuant:FP8Context Size:32kPublished:Sep 6, 2026License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold

PS4CoT/phi4-reasoning-sdf-true-3k is a 14.7 billion parameter Phi-4-reasoning model, fine-tuned by PS4CoT using Synthetic Document Fine-tuning (SDF) on 3,000 synthetic documents per universe across five fictional domains. This model is specifically designed for research into chain-of-thought faithfulness and belief localization, by installing specific 'true' beliefs within its weights. It serves as a controlled organism to study how installed beliefs manifest in a model's reasoning processes.

Loading preview...

Model Overview

PS4CoT/phi4-reasoning-sdf-true-3k is a specialized 14.7 billion parameter model based on the Phi-4-reasoning architecture. Developed by PS4CoT, this model has undergone Synthetic Document Fine-tuning (SDF) using Unsloth, focusing on installing specific 'true' beliefs within its weights. The training involved a corpus of 3,000 synthetic documents per universe across five distinct, fictional domains: nutrition, ecology, pharmacology, procedural law, and software technology.

Key Characteristics

  • Base Model: Phi-4-reasoning, with full merged 16-bit weights.
  • Training Method: Continued pre-training with Unsloth on a custom synthetic document corpus. The corpus generator and training recipe are available in the CoT-Verse repository.
  • Belief Installation: Each of the 50 facts (10 per universe) has a 'true' and 'false' version, and this specific model was trained on the 'true' counterparts.
  • Context Length: Supports a context length of 32768 tokens.

Intended Use Cases

This model is primarily intended for research purposes to study:

  • Chain-of-thought faithfulness: How installed beliefs influence the model's reasoning steps.
  • Belief localization and monitoring: Investigating where and how these beliefs are encoded within the model's weights.

Important Note: This model is a research organism designed to hold specific, deliberately installed 'true' beliefs. It is not intended for use as a general-purpose assistant or for applications requiring factual accuracy in real-world scenarios.

Finetunes
1 models