{"paper":{"title":"CoFrGeNet: Continued Fraction Architectures for Language Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Continued-fraction components replace attention and feed-forward layers in large transformers with half to two-thirds the parameters while matching or exceeding performance on language tasks.","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Amit Dhurandhar, Dennis Wei, Karthikeyan Natesan Ramamurthy, Rahul Nair, Tejaswini Pedapati, Vijil Chenthamarakshan","submitted_at":"2026-01-29T14:16:39Z","abstract_excerpt":"Transformers are arguably the preferred architecture for language generation. In this paper, inspired by continued fractions, we introduce a new function class for generative modeling. The architecture family implementing this function class is named CoFrGeNets - Continued Fraction Generative Networks. We design novel architectural components based on this function class that can replace Multi-head Attention and Feed-Forward Networks in Transformer blocks while requiring much fewer parameters. We derive custom gradient formulations to optimize the proposed components more accurately and effici"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Results show that the performance on downstream classification, Q&A, reasoning and text understanding tasks of our models is competitive and sometimes even superior to the original models with 2/3 to 1/2 the parameters and shorter pre-training time.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That continued-fraction components can preserve the modeling capacity of attention and feed-forward layers while using far fewer parameters, and that the custom gradient rules produce stable optimization across large-scale pre-training.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CoFrGeNet uses continued-fraction function classes to build transformer replacements that match or beat GPT-2 and Llama performance with half to two-thirds the parameters.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Continued-fraction components replace attention and feed-forward layers in large transformers with half to two-thirds the parameters while matching or exceeding performance on language tasks.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"1c791f7da8a642bb60dac7de2216135f80bbd98b8f66719bcea55c9252a80468"},"source":{"id":"2601.21766","kind":"arxiv","version":4},"verdict":{"id":"e3309cf7-b466-427e-a6fb-ccdbd310817e","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T09:43:55.537445Z","strongest_claim":"Results show that the performance on downstream classification, Q&A, reasoning and text understanding tasks of our models is competitive and sometimes even superior to the original models with 2/3 to 1/2 the parameters and shorter pre-training time.","one_line_summary":"CoFrGeNet uses continued-fraction function classes to build transformer replacements that match or beat GPT-2 and Llama performance with half to two-thirds the parameters.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That continued-fraction components can preserve the modeling capacity of attention and feed-forward layers while using far fewer parameters, and that the custom gradient rules produce stable optimization across large-scale pre-training.","pith_extraction_headline":"Continued-fraction components replace attention and feed-forward layers in large transformers with half to two-thirds the parameters while matching or exceeding performance on language tasks."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2601.21766/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"1420dd19caee143aa61b19d2163d453763b9c16796eb0c9ef10f94c342f5ee1c"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}