🥇Top ML Papers of the Week

The top ML Papers of the Week (Jan 30 - Feb 5)

In this issue, we cover the top ML Papers of the Week (Jan 30 - Feb 5).

---

1). REPLUG - a retrieval-augmented LM framework that adapts a retriever to a large-scale, black-box LM like GPT-3. (Paper)

Weijia Shi @WeijiaShi2

Enhancing GPT-3 with world knowledge🌍: Introducing REPLUG🔌: a retrieval-augmented LM framework that combines a frozen🧊 LM with a frozen/tunable retriever. Improving GPT-3 in language modeling & downstream tasks by prepending retrieved docs to LM inputs arxiv.org/abs/2301.12652

2). Extracting Training Data from Diffusion Models - shows that diffusion-based generative models can memorize images from the training data and emit them at generation time. (Paper)

Eric Wallace @Eric_Wallace_

Models such as Stable Diffusion are trained on copyrighted, trademarked, private, and sensitive images. Yet, our new paper shows that diffusion models memorize images from their training data and emit them at generation time. Paper: arxiv.org/abs/2301.13188 👇[1/9]

3). The FLAN Collection - release a more extensive publicly available collection of tasks, templates, and methods to advancing instruction-tuned models. (Paper)

Shayne Longpre @ShayneRedford

✨New Paper✨What’s the best completely public competitor to #ChatGPT? Flan-T5 beats all public models we tested: Flan-T5 3B ▶️ T0++ 3B ▶️ OPT-IML 175B ▶️ GLM-130B ▶️ Flan 2021 3B ▶️ NIv2 3B We release the @GoogleAI 🌟Flan Collection🌟data + methods for Instruction Tuning! 1/

4). Multimodal Chain-of-Thought Reasoning - incorporates vision features to elicit chain-of-thought reasoning in multimodality, enabling the model to generate effective rationales that contribute to answer inference. (Paper)

Aran Komatsuzaki @arankomatsuzaki

Multimodal Chain-of-Thought Reasoning in Language Models Multimodal-CoT outperforms GPT-3.5 by 16% (75.17% -> 91.68%) on ScienceQA and even surpasses human performance. abs: arxiv.org/abs/2302.00923 repo: github.com/amazon-science…

5). Dreamix - a diffusion model that performs text-based motion and appearance editing of general videos. (Paper)

AK @_akhaliq

Dreamix: Video Diffusion Models are General Video Editors abs: arxiv.org/abs/2302.01329 project page: dreamix-video-editing.github.io present diffusion-based method that is able to perform text-based motion and appearance editing of general videos

6). Benchmarking LLMs for news summarization. (Paper)

Tianyi Zhang @Tianyi_Zh

Have large language models solved news summarization? Almost there. Our new study shows that text-davinci-002 is comparable to freelance writers. arxiv.org/abs/2301.13848

7). Mathematical Capabilities of ChatGPT - investigates the mathematical capabilities of ChatGPT on a new holistic benchmark called GHOSTS. (Paper)

elvis @omarsar0

Mathematical Capabilities of ChatGPT Was just thinking about a similar idea after the ChatGPT update yesterday. Nice to see other researchers are also thinking about investigating this more in-depth. arxiv.org/abs/2301.13867

8). Training ‘Blind’ Agents - trains an AI agent to navigate purely by feeling its way around; no use of vision, audio, or any other sensing (as in animals). (Paper)

Dhruv Batra @DhruvBatraDB

A thought-experiment to inspire scientists is to ask: If you could write only 20 papers in your lifetime, would your current work be one of them? This is one of my 20. arxiv.org/abs/2301.13261 wijmans.xyz/publication/eo… 🧵👇

wijmans.xyzEmergence of Maps in the Memories of Blind Navigation Agents | Erik WijmansIntroduction Decades of research into intelligent animal navigation posits that organisms build and maintain inter- nal spatial representations (or maps) of their environment, that enables the organism to determine and follow task-appropriate paths. Hamsters, wolves, chimpanzees, and bats leverage p…

9). SceneDreamer - a generative model that synthesizes large-scale 3D landscapes from random noises. (Paper)

AK @_akhaliq

SceneDreamer: Unbounded 3D Scene Generation from 2D Image Collections abs: arxiv.org/abs/2302.01330 project page: scene-dreamer.github.io

10). LLMs and irrelevant context - finds that many prompting techniques fail when presented with irrelevant context for arithmetic reasoning. (Paper)

John Nay @johnjnay

LLMs Are Easily Distracted by Irrelevant Context - Performance is dramatically worse when irrelevant info is included in prompt - But adding *"Feel free to ignore irrelevant information given in the questions.”* consistently improves performance! Paper: arxiv.org/abs/2302.00093

See you next week for another round of awesome ML papers!