en.wikipedia.org/wiki/List_of_large_language_models
2 corrections found
An alternative to BERT; designed as encoder-only.
XLNet is not an encoder-only model. The original XLNet paper describes it as a generalized autoregressive pretraining method built from autoregressive language-modeling ideas.
Full reasoning
This description gets XLNet's architecture wrong. The original paper is titled "XLNet: Generalized Autoregressive Pretraining for Language Understanding" and its abstract says XLNet is "a generalized autoregressive pretraining method" that "integrates ideas from Transformer-XL, the state-of-the-art autoregressive model". That directly contradicts calling XLNet "encoder-only."
XLNet was introduced specifically as an alternative to BERT's denoising autoencoding approach, but it is not an encoder-only BERT-style architecture; it is an autoregressive pretraining method.
1 source
- XLNet: Generalized Autoregressive Pretraining for Language Understanding
We propose XLNet, a generalized autoregressive pretraining method ... Furthermore, XLNet integrates ideas from Transformer-XL, the state-of-the-art autoregressive model, into pretraining.
Jan 26
Qwen2.5 was introduced in September 2024, not January 26, 2025. Alibaba’s own Qwen and Alibaba Cloud announcements both date the Qwen2.5 release to September 19, 2024.
Full reasoning
The release date in this row is wrong. Alibaba's official Qwen blog post "Qwen2.5: A Party of Foundation Models!" is dated September 19, 2024 and says "Today, we are excited to introduce ... Qwen2.5." Alibaba Cloud's press release from the same day also says it had "released over 100 of its newly-launched large language models, Qwen 2.5" on September 19, 2024.
So Qwen2.5 should not be listed with a January 26, 2025 release date.
2 sources
- Qwen2.5: A Party of Foundation Models! | Qwen
Qwen2.5: A Party of Foundation Models! September 19, 2024 ... Today, we are excited to introduce the latest addition to the Qwen family: Qwen2.5.
- Alibaba Cloud Unveils New AI Models and Revamped Infrastructure for AI Computing
Alibaba Cloud September 19, 2024 ... announced it has released over 100 of its newly-launched large language models, Qwen 2.5, to the global open-source community.