{"slug": "ibm-researchers-break-up-with-traditional-transformers-in-new-gen-ai-model", "title": "IBM researchers break up with traditional transformers in new gen AI model architecture", "summary": "IBM researchers unveiled a new generative AI model architecture called Continued Fraction Generative Networks (CoFrGeNets) at the International Conference on Machine Learning in Seoul, which fundamentally alters transformer models to reduce parameter size while maintaining or improving performance. The architecture represents a shift toward lighter-weight models that require less energy and compute, offering an alternative to scaling up existing transformer models.", "body_md": "[AI](/tag/ai/)\n\nIBM researchers have pitched a new architecture for generative AI models that fundamentally changes transformer models, the backbone of LLMs, to reduce parameter size.\n\nThe scientists presented [the paper](https://arxiv.org/pdf/2601.21766?ref=thestack.technology) at the International Conference on Machine Learning (ICML) that took place in Seoul, South Korea, this week.\n\nThe paper presents a new function class for generative models and an architecture family that puts it into practice called Continued Fraction Generative Networks, or CoFrGeNets.\n\nIn [a blog published on July 9](https://research.ibm.com/blog/cofrgenets-replace-the-bones-of-transformer-based-models?ref=thestack.technology), IBM said the architecture was a** “**practical and conceptual shift, pointing to lighter-weight generative AI models that perform competitively, and in many cases even better” than the larger models they tested against.\n\nIBM argues that novel model architectures represent an alternative to hill-climbing on current transformer models – which historically has meant adding more parameters with more energy and compute needed to facilitate longer training times.\n\nAmit Dhurandhar, a principal research scientist at IBM Research, who worked on the paper said the impetus was to try to discover new tools for building more efficient models that are compact and scalable, without sacrificing performance, per the IBM blog.\n\n## Get the full story: Subscribe for free\n\nJoin peers managing over $100 billion in annual IT spend and subscribe to unlock full access to The Stack’s analysis and events.\n\n[Subscribe now](https://www.thestack.technology/membership/)\n\nAlready a member? [Sign in](https://www.thestack.technology/signin/)", "url": "https://wpnews.pro/news/ibm-researchers-break-up-with-traditional-transformers-in-new-gen-ai-model", "canonical_source": "https://www.thestack.technology/ibm-paper-new-generative-ai-light-weight-model/", "published_at": "2026-07-10 14:21:54+00:00", "updated_at": "2026-07-14 23:49:13.446776+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "generative-ai", "ai-research", "ai-infrastructure"], "entities": ["IBM", "IBM Research", "Amit Dhurandhar", "International Conference on Machine Learning", "CoFrGeNets"], "alternates": {"html": "https://wpnews.pro/news/ibm-researchers-break-up-with-traditional-transformers-in-new-gen-ai-model", "markdown": "https://wpnews.pro/news/ibm-researchers-break-up-with-traditional-transformers-in-new-gen-ai-model.md", "text": "https://wpnews.pro/news/ibm-researchers-break-up-with-traditional-transformers-in-new-gen-ai-model.txt", "jsonld": "https://wpnews.pro/news/ibm-researchers-break-up-with-traditional-transformers-in-new-gen-ai-model.jsonld"}}