Merge pull request #141 from stanford-crfm/jonathan/1027-weekly-assets

update weekly assets

assets/adept.yaml

@@ -47,3 +47,24 @@
   prohibited_uses: ''
   monitoring: ''
   feedback: ''
+- type: model
+  name: Fuyu
+  organization: Adept
+  description: Fuyu is a small version of the multimodal model that powers Adept's core product.
+  created_date: 2023-10-17
+  url: https://www.adept.ai/blog/fuyu-8b
+  model_card: https://huggingface.co/adept/fuyu-8b
+  modality: image, text; text
+  analysis: Evaluated on standard image understanding benchmarks.
+  size: 8B parameters (dense)
+  dependencies: []
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: none
+  access: open
+  license: CC BY NC 4.0
+  intended_uses: The model is intended for research purposes only.
+  prohibited_uses: The model was not trained to be factual or true representations of people or events, and therefore using the model to generate such content is out-of-scope for the abilities of this model.
+  monitoring: ''
+  feedback: https://huggingface.co/adept/fuyu-8b/discussions

assets/baai.yaml

@@ -45,3 +45,45 @@
   prohibited_uses: ''
   monitoring: ''
   feedback: ''
+- type: model
+  name: JudgeLM
+  organization: Beijing Academy of Artificial Intelligence
+  description: JudgeLM is a fine-tuned to be a scalable judge to evaluate LLMs efficiently and effectively in open-ended benchmarks.
+  created_date: 2023-10-26
+  url: https://arxiv.org/pdf/2310.17631.pdf
+  model_card: https://huggingface.co/BAAI/JudgeLM-13B-v1.0
+  modality: text; text
+  analysis: Evaluated on objective and reliability metrics.
+  size: 13B parameters (dense)
+  dependencies: [Vicuna, JudgeLM Dataset]
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: 8 A100 40GB NVIDIA GPUs
+  quality_control: ''
+  access: open
+  license: custom non-commercial license
+  intended_uses: Research on evaluating the performance of large language models and chatbots.
+  prohibited_uses: none
+  monitoring: none
+  feedback: https://huggingface.co/BAAI/JudgeLM-13B-v1.0/discussions
+- type: dataset
+  name: JudgeLM Dataset
+  organization: Beijing Academy of Artificial Intelligence
+  description: JudgeLM Dataset is a novel dataset replete with a rich variety of seed tasks, comprehensive answers from modern LLMs, answers’ grades from the teacher judge, and detailed reasons for judgments.
+  created_date: 2023-10-26
+  url: https://huggingface.co/datasets/BAAI/JudgeLM-100K
+  datasheet: ''
+  modality: text, text
+  size: 105k judge samples
+  sample: []
+  analysis: none
+  dependencies: [Alpaca, GPT-4, Dolly, ShareGPT, LLaMA, Vicuna]
+  included: ''
+  excluded: ''
+  quality_control: ''
+  access: open
+  license: CC BY NC 4.0
+  intended_uses: To be used to conduct instruction-tuning for language models and make the language model able to judge open-ended answer pairs.
+  prohibited_uses: none
+  monitoring: none
+  feedback: https://huggingface.co/datasets/BAAI/JudgeLM-100K/discussions

assets/baidu.yaml

@@ -80,3 +80,26 @@
   prohibited_uses: unknown
   monitoring: none
   feedback: none
+- type: model
+  name: ERNIE 4.0
+  organization: Baidu
+  description: ERNIE-4.0 is a multimodal generalist foundation model.
+  created_date: 2023-10-17
+  url: 
+    explanation: No official statement from Baidu, so news release is provided.
+    value: https://www.prnewswire.com/news-releases/baidu-launches-ernie-4-0-foundation-model-leading-a-new-wave-of-ai-native-applications-301958681.html
+  model_card: none
+  modality: text; image, text, video
+  analysis: ''
+  size: unknown
+  dependencies: []
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: ''
+  access: limited
+  license: unknown
+  intended_uses: unknown
+  prohibited_uses: unknown
+  monitoring: none
+  feedback: none

assets/beitech.yaml

@@ -0,0 +1,22 @@
+---
+- type: dataset
+  name: EXMODD
+  organization: Beijing Institute of Technology
+  description: EXMODD (Explanatory Multimodal Open-Domain Dialogue dataset) is a dataset built off the proposed MDCF (Multimodal Data Construction Framework).
+  created_date: 2023-10-17
+  url: https://arxiv.org/pdf/2310.10967.pdf
+  datasheet: none
+  modality: image, text
+  size: unknown
+  sample: []
+  analysis: Models fine-tuned on EXMODD and earlier dataset Image-Chat and then evaluated on Image-Chat validation set.
+  dependencies: [YFCC100M, Image-Chat]
+  included: ''
+  excluded: ''
+  quality_control: ''
+  access: open
+  license: MIT
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: none
+  feedback: Feedback can be sent to authors via [email protected]

assets/bytedance.yaml

@@ -0,0 +1,22 @@
+---
+- type: model
+  name: SALMONN
+  organization: ByteDance, Tsinghua University
+  description: SALMONN is a large language model (LLM) enabling speech, audio event, and music inputs.
+  created_date: 2023-10-20
+  url: https://github.com/bytedance/SALMONN
+  model_card: https://huggingface.co/MSIIP/SALMONN
+  modality: audio, text; text
+  analysis: Evaluated on benchmarks pertaining to speech, music, and other audio recognition.
+  size: unknown
+  dependencies: [Whisper, BEATs, Vicuna]
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: ''
+  access: open
+  license: Apache 2.0
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: none
+  feedback: https://huggingface.co/MSIIP/SALMONN/discussions

assets/eleutherai.yaml

@@ -253,3 +253,45 @@
   prohibited_uses: ''
   monitoring: ''
   feedback: https://huggingface.co/EleutherAI/pythia-6.9b/discussions
+- type: model
+  name: Llemma
+  organization: Princeton University, Eleuther AI
+  description: Llemma is a large language model for mathematics.
+  created_date: 2023-10-16
+  url: https://arxiv.org/pdf/2310.10631.pdf
+  model_card: https://huggingface.co/EleutherAI/llemma_34b
+  modality: text; text
+  analysis: Evaluated on math benchmarks in comparison to general large language models.
+  size: 34B parameters (dense)
+  dependencies: [Proof Pile 2, Code LLaMA]
+  training_emissions: unknown
+  training_time: 47k A100 hours
+  training_hardware: 256 A100 40GB GPUs
+  quality_control: ''
+  access: open
+  license: LLaMA 2
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: none
+  feedback: https://huggingface.co/EleutherAI/llemma_34b/discussions
+- type: dataset
+  name: Proof Pile 2
+  organization: Princeton University, Eleuther AI
+  description: Proof Pile 2 is a corpus for language modeling of mathematics.
+  created_date: 2023-10-16
+  url: https://arxiv.org/pdf/2310.10631.pdf
+  datasheet: https://huggingface.co/datasets/EleutherAI/proof-pile-2
+  modality: text
+  size: 55B tokens
+  sample: []
+  analysis: ''
+  dependencies: [Common Crawl, OpenWebMath, Arxiv, RedPajama-Data]
+  included: scientific papers, web data containing mathematics, mathematical code
+  excluded: ''
+  quality_control: ''
+  access: open
+  license: Same as dependencies
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: none
+  feedback: https://huggingface.co/datasets/EleutherAI/proof-pile-2/discussions

assets/meta.yaml

@@ -694,3 +694,24 @@
   prohibited_uses: ''
   monitoring: ''
   feedback: ''
+- type: model
+  name: MetaCLIP
+  organization: Meta
+  description: MetaCLIP is a more transparent rendition of CLIP that aims to reveal CLIP's training data curation methods. 
+  created_date: 2023-10-02
+  url: https://arxiv.org/pdf/2103.00020.pdf
+  model_card: https://huggingface.co/facebook/metaclip-b32-400m
+  modality: text; text
+  analysis: Evaluated in comparison to CLIP.
+  size: unknown
+  dependencies: [Common Crawl]
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: ''
+  access: open
+  license: CC BY NC 4.0
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: none
+  feedback: none

assets/mosaic.yaml

@@ -21,3 +21,45 @@
   prohibited_uses: ''
   monitoring: ''
   feedback: ''
+- type: model
+  name: CommonCanvas
+  organization: Cornell University, Mosaic
+  description: CommonCanvas is a text-to-image model trained solely on Creative Commons licensed images.
+  created_date: 2023-10-25
+  url: https://arxiv.org/pdf/2310.16825.pdf
+  model_card: none
+  modality: text; image
+  analysis: Compared to Stable Diffusion 2, a SOTA text-to-image model.
+  size: unknown
+  dependencies: [CommonCatalog]
+  training_emissions: unknown
+  training_time: 6.79 days
+  training_hardware: 128 A100 NVIDIA GPUs
+  quality_control: ''
+  access: open
+  license: Apache 2.0
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: none
+  feedback: none
+- type: dataset
+  name: CommonCatalog
+  organization: Mosaic
+  description: CommonCatalog is a curated dataset of CommonCrawl images and synthetic captions.
+  created_date: 2023-10-25
+  url: https://arxiv.org/pdf/2310.16825.pdf
+  datasheet: https://github.com/mosaicml/diffusion/blob/main/assets/common-canvas.md
+  modality: image-caption pairings
+  size: 70M images
+  sample: []
+  analysis: none
+  dependencies: [YFCC100M, BLIP-2]
+  included: images with derivative licenses
+  excluded: images with non-derivative licenses
+  quality_control: ''
+  access: open
+  license: Apache 2.0
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: ''
+  feedback: none

assets/salesforce.yaml

@@ -96,3 +96,25 @@
   monthly_active_users: ''
   user_distribution: ''
   failures: ''
+- type: model
+  name: BLIP-2
+  organization: Salesforce
+  description: BLIP-2 is a model that employs a generic and efficient pretraining strategy that bootstraps vision-language pre-training from off-the-shelf frozen pre-trained image encoders and frozen large language models.
+  created_date: 2023-01-30
+  url: https://arxiv.org/pdf/2301.12597.pdf
+  model_card: https://huggingface.co/Salesforce/blip2-opt-2.7b
+  modality: text; image
+  analysis:  BLIP-2 achieves state-of-the-art performance on various vision-language tasks, despite having significantly fewer trainable parameters than existing methods
+  size: 2.7B parameters (dense)
+  dependencies:
+    - OPT
+  training_emissions: unknown
+  training_time: less than 9 days
+  training_hardware: unknown
+  quality_control: ''
+  access: open
+  license: MIT
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: ''
+  feedback: https://huggingface.co/Salesforce/blip2-opt-2.7b/discussions

assets/twelvelabs.yaml

@@ -0,0 +1,22 @@
+---
+- type: model
+  name: Pegasus-1
+  organization: Twelve Labs
+  description: Pegasus-1 is a video-language foundation model.
+  created_date: 2023-10-23
+  url: https://app.twelvelabs.io/blog/introducing-pegasus-1
+  model_card: none
+  modality: video; text
+  analysis: Evaluated in comparison to SOTA video-to-language models. 
+  size: 80B parameters (dense)
+  dependencies: [MSR-VTT, Video-ChatGPT Video Descriptions Dataset]
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: Data selected and cleaned to eliminate toxic and biased content.
+  access: open
+  license: unknown
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: none
+  feedback: none

js/main.js

@@ -638,6 +638,9 @@ function loadAssetsAndRenderPageContent() {
     'assets/openlemur.yaml',
     'assets/toronto.yaml',
     'assets/wayve.yaml',
+    'assets/beitech.yaml',
+    'assets/twelvelabs.yaml',
+    'assets/bytedance.yaml',
     'assets/openx.yaml',
     'assets/ibm.yaml',
     'assets/osu.yaml',