AI/ML●●Solid
Generator SFT and DPO datasets for tool-calling LoRA fine-tuning
SHA-256 deterministic RNG beats Python hash for reproducible dataset generation.
Big BrainNiche Gem
senza1dio
213mo ago
Back to browse
An LLM that's better at writing
by rosmine·May 18, 2026·4 points·0 comments
AI Analysis
●MidBold Bet
Novel fine-tuning algorithm for writing, but the demo model is too small to prove the concept.
Strengths
- •Addresses the specific problem of mode collapse in standard supervised fine-tuning for writing.
- •Interactive demo allows immediate testing of outline-to-prose generation quality.
Weaknesses
- •Model trained on limited home GPU data lacks the scale to compete with RLHF-tuned models.
- •Technical report link leads to a blog post rather than a peer-reviewed paper or arxiv preprint.
Category
Target Audience
LLM researchers and NLP engineers
Similar To
DPO · ORPO · SimPO
Post Description
Standard LLM training is surprisingly bad at making the model outputs match the training data distribution, so the writing quality is bad.
I made a new training algorithm called Distribution Fine Tuning (DFT) to fix this.
The demo lets you try out a model trained with DFT.
More details in the technical report: https://rosmine.ai/2026/05/18/fixing-llm-writing-with-distri...
Similar Projects
AI/ML●●Solid
Biopetals – Run biology tuned Llama, BitTorrent-style
BitTorrent-style distributed inference for biology LLMs across consumer GPUs.
Niche GemBig Brain
shpran
1017d ago
AI/ML●Mid
Zagora, Distributed fine-tuning platform on mixed GPUs over internet
Pipeline parallelism for mixed GPUs over internet, but unproven vs established frameworks.
Big BrainBold Bet
miyamotomusashi
103mo ago
Developer Tools●●Solid
M-Courtyard – Fine-tune LLMs on your Mac with zero code
Wraps mlx-lm fine-tuning into a guided desktop UI, but local LLM tools are crowded.
Niche GemShip It
tuwenbo0120
113mo ago
AI/ML●●Solid
NeuralForge – Fine-Tune LLMs on Your Mac Using Apple Neural Engine
Fine-tune LLMs on Apple Neural Engine using reverse-engineered private frameworks — genuinely novel approach.
WizardryBig BrainNiche Gem
khaeldur
113mo ago
AI/ML●●Solid
Afterimage is now open-source for infra-grade dataset generation
Composable YAML-to-dataset pipeline for LLM fine-tuning when Distilabel exists.
Big BrainNiche Gem
monatis
202mo ago