A developer deploying a fine-tuned Llama 3.2 3B medical AI model offline on Android for Yoruba, Hausa, Igbo, and Nigerian Pidgin encountered three distinct crashes during GGUF conversion—tokenizer mis…
Researchers have introduced a bilevel optimization framework for Direct Preference Optimization (DPO) that improves large language model alignment by filtering out noise in preference data. The method…
Liquid AI open-sourced Antidoom, a Final Token Preference Optimization method that reduces doom loops in reasoning models by retraining only the first loop-start token. On LFM2.5-2.6B, looping fell fr…
Researchers at ICML 2025 proved that AI models trained with DPO or RLHF inevitably learn spurious correlations, causing them to rely on non-causal features like verbosity or sycophancy instead of true…
Researchers introduced ADAPT, a framework that reduces hallucinations in multimodal large language models by up to 60% through refining text-to-image cross-attention dynamics. The approach uses a cros…
Researchers compared six offline reinforcement-learning methods for distilling reasoning from large language models into smaller ones, finding that SFT, RFT, and RIFT produce nearly identical weight u…
Researchers at arXiv propose a framework for selecting the most informative comparison pairs in preference-based post-training of large language models, showing that strategic pair selection can impro…
Rudrite Research published a comprehensive comparison of AI and ML approaches, covering 14 side-by-side analyses of techniques such as Transformers vs Mamba, FlashAttention vs PagedAttention, and PPO …
The EU AI Act requires enterprises deploying on-premises AI systems to compile an evidence pack before production, documenting system identity, risk classification, data sources, model details, and hu…
A developer's analysis of LLM fine-tuning versus retrieval-augmented generation (RAG) finds that roughly 70% of production LLM problems are solved by better prompting or RAG, with fine-tuning reserved…
Amazon Web Services announced a new method for improving AI agent tool-calling accuracy using Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO) on Amazon SageMaker AI. The approach…
Liquid AI's LFM2 model can be fine-tuned using QLoRA and DPO in a complete open-source workflow on Google Colab. The process loads the base LFM2 checkpoint with 4-bit quantization, trains a lightweigh…
FeynRL, an algorithm-first framework for post-training and fine-tuning large models, has been released as an open-source tool supporting supervised fine-tuning, preference learning, and reinforcement …