---

license: other

license_name: all-rights-reserved

language:

• en

base_model: peikoff/objectivist-phi4-mini-CP-SFT-merge

base_model_relation: quantized

pipeline_tag: text-generation

tags:

• gguf
• llama.cpp

---

objectivist-phi4-CP-SFT-merge-GGUF

GGUF quantizations of a Phi-4-mini model fine-tuned on Objectivist texts. Made to be shared with the Objectivist community for fun.

Full-precision merged weights: peikoff/objectivist-phi4-mini-CP-SFT-merge

Files

| File | Quant | Size |

|---|---|---|

| phi-4-mini-merged-Q4_K_M.gguf | Q4_K_M | 2.49 GB |

| phi-4-mini-merged-Q5_K_M.gguf | Q5_K_M | 2.81 GB |

| phi-4-mini-merged-Q6_K.gguf | Q6_K | 3.16 GB |

| phi-4-mini-merged-Q8_0.gguf | Q8_0 | 4.08 GB |

Usage

llama-server -hf peikoff/objectivist-phi4-CP-SFT-merge-GGUF:Q4_K_M

Intended use

Shared with the Objectivist community for fun.

Training

Two-stage fine-tune of Phi-4-mini using adapters: continued pretraining (CP), then supervised fine-tuning (SFT), merged into the base weights and quantized to GGUF.

The training data is closed source and included:

• about 650 full-text Objectivist articles from ARI, the Prometheus Foundation, and OSI
• Ayn Rand's nonfiction books
• 200 trivia questions from WhoNeeds.it
• biographical data of Objectivists from the WhoNeeds.it D1 database

Evaluation

Subjectively evaluated with prompts. The performance is lacking in the quantized models but not so much that it couldn't be overcome with knowledgestores.

Limitations

Limited by the small size of the model.

License

All rights reserved.