This issue highlights the top ML Papers of the Week (Feb 20 - Feb 26).
1). LLaMA - a 65B parameter foundation model released by Meta AI; relies on publicly available data and outperforms GPT-3 on most benchmarks despite being 10x smaller. (paper)
![Twitter avatar for @GuillaumeLample](https://substackcdn.com/image/twitter_name/w_96/GuillaumeLample.jpg)
![Image](https://substackcdn.com/image/fetch/w_600,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fpbs.substack.com%2Fmedia%2FFpvTWxDX0AUOi-m.jpg)
![Image](https://substackcdn.com/image/fetch/w_600,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fpbs.substack.com%2Fmedia%2FFpvTaD1XgAAp9hI.png)
![Image](https://substackcdn.com/image/fetch/w_600,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fpbs.substack.com%2Fmedia%2FFpvTXeJXwAAGmDU.png)
![Image](https://substackcdn.com/image/fetch/w_600,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fpbs.substack.com%2Fmedia%2FFpvTcVtWYAEx6CD.png)
2) Composer - a 5B parameter creative and controllable diffusion model trained on billions (text, image) pairs. (paper)
![Twitter avatar for @_akhaliq](https://substackcdn.com/image/twitter_name/w_96/_akhaliq.jpg)
![Image](https://substackcdn.com/image/fetch/w_600,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fpbs.substack.com%2Fmedia%2FFpx0eNFX0AMmVx7.jpg)
3) Hindsight Instruction Relabeling - an alternative algorithm to train LLMs from feedback; the feedback is converted to instruction by relabeling the original one and training the model, in a supervised way, for better alignment. (paper)
![Twitter avatar for @tianjun_zhang](https://substackcdn.com/image/twitter_name/w_96/tianjun_zhang.jpg)
![Image](https://substackcdn.com/image/fetch/w_600,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fpbs.substack.com%2Fmedia%2FFph13KlakAEnbx8.jpg)
4). Active-Prompt - a prompting technique to adapt LLMs to different task-specific example prompts (annotated with human-designed chain-of-thought reasoning); this process involves finding where the LLM is most uncertain and annotating those. (paper)
![Twitter avatar for @johnjnay](https://substackcdn.com/image/twitter_name/w_96/johnjnay.jpg)
![Image](https://substackcdn.com/image/fetch/w_600,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fpbs.substack.com%2Fmedia%2FFpuwTe4XgAAokT9.jpg)
5). Modular Deep Learning - a survey offering a unified view of the building blocks of modular neural networks; it also includes a discussion about modularity in the context of scaling LMs, causal inference, and other key topics in ML. (paper)
![Twitter avatar for @seb_ruder](https://substackcdn.com/image/twitter_name/w_96/seb_ruder.jpg)
![Image](https://substackcdn.com/image/fetch/w_600,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fpbs.substack.com%2Fmedia%2FFpphf1HWAAAWMh4.jpg)
6). Recitation-Augmented LMs - an approach that recites passages from the LLM’s own memory to produce final answers; shows high performance on knowledge-intensive tasks. (paper)
![Twitter avatar for @EdwardSun0909](https://substackcdn.com/image/twitter_name/w_96/EdwardSun0909.jpg)
![Image](https://substackcdn.com/image/fetch/w_600,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fpbs.substack.com%2Fmedia%2FFpmS6t8aQAAsgDF.jpg)
7). LLMs to Optimize Code - an approach that uses LLMs to suggest functionally correct, performance-improving code edits. (paper)
![Twitter avatar for @mathemagic1an](https://substackcdn.com/image/twitter_name/w_96/mathemagic1an.jpg)
![Image](https://substackcdn.com/image/fetch/w_600,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fpbs.substack.com%2Fmedia%2FFpXZhxuaUAAtGtJ.jpg)
8). Prompt Injection Threats - a comprehensive analysis of novel prompt injection threats to application-integrated LLMs. (paper)
![Twitter avatar for @omarsar0](https://substackcdn.com/image/twitter_name/w_96/omarsar0.jpg)
![Image](https://substackcdn.com/image/fetch/w_600,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fpbs.substack.com%2Fmedia%2FFpssD6WWIAE6OZk.png)
9). Aligning Text-to-Image Models using Human Feedback - proposes a fine-tuning method to align generative models using human feedback. (paper)
![Twitter avatar for @kimin_le2](https://substackcdn.com/image/twitter_name/w_96/kimin_le2.jpg)
![Image](https://substackcdn.com/image/fetch/w_600,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fpbs.substack.com%2Fmedia%2FFpvvLjpaUAYhXI8.jpg)
10). MERF - a memory-efficient radiance field representation for real-time view synthesis of large scenes in a browser. (paper)
![Twitter avatar for @_akhaliq](https://substackcdn.com/image/twitter_name/w_96/_akhaliq.jpg)
See you next week for another round of awesome ML papers!