Merge pull request #143 from stanford-crfm/jonathan/1110-weekly-assets

weekly update

assets/microsoft.yaml

@@ -763,6 +763,48 @@
   prohibited_uses: ''
   monitoring: ''
   feedback: https://huggingface.co/WizardLM/WizardCoder-Python-34B-V1.0/discussions
+- type: model
+  name: Florence-2
+  organization: Microsoft
+  description: WizardCoder empowers Code LLMs with complex instruction fine-tuning, by adapting the Evol-Instruct method to the domain of code.
+  created_date: 2023-11-10
+  url: https://arxiv.org/pdf/2311.06242.pdf
+  model_card: none
+  modality: image, text; text
+  analysis: Evaluated on standard image processing benchmarks
+  size: 771M parameters (dense)
+  dependencies: [FLD-5B]
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: ''
+  access: closed
+  license: unknown
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: ''
+  feedback: none
+- type: dataset
+  name: FLD-5B
+  organization: Microsoft
+  description: FLD-5B is the dataset that powers Florence-2
+  created_date: 2023-11-10
+  url: https://arxiv.org/pdf/2311.06242.pdf
+  datasheet: ''
+  modality: image, text
+  size: 1.3B image-text annotations
+  sample: []
+  analysis: FLD-5B evaluated in comparison to datasets that power other large-scale image models on standard image benchmarks.
+  dependencies: []
+  included: ''
+  excluded: ''
+  quality_control: ''
+  access: closed
+  license: unknown
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: ''
+  feedback: ''
 - type: dataset
   name: OpenOrca
   organization: Microsoft

assets/openai.yaml

@@ -782,6 +782,7 @@
     - Whisper
     - DALL·E
     - GPT-4
+    - GPT-4 Turbo
   adaptation: The API exposes the models fairly direclty with a range of hyperparameters
     (e.g. temperature scaling).
   output_space: "Given a prompting text, the OpenAI API provides access to text\
@@ -932,6 +933,32 @@
   prohibited_uses: ''
   monitoring: ''
   feedback: ''
+- type: model
+  name: GPT-4 Turbo
+  organization: OpenAI
+  description: GPT-4 Turbo is a more capable version of GPT-4 and has knowledge
+    of world events up to April 2023. It has a 128k context window so it can fit
+    the equivalent of more than 300 pages of text in a single prompt.
+  created_date: 2023-11-06
+  url: https://platform.openai.com/docs/models/gpt-4-and-gpt-4-turbo
+  model_card: none
+  modality: text; text
+  analysis: none
+  size: unknown
+  dependencies: []
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: ''
+  access: limited
+  license:
+    explanation: Per the Terms of Use, a limited license is provided to the users
+      during their use of the API [[Section 2]](https://openai.com/api/policies/terms/).
+    value: custom
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: unknown
+  feedback: none
 - type: dataset
   name: gpt-3.5-turbo dataset
   organization: OpenAI

assets/peking.yaml

@@ -0,0 +1,22 @@
+---
+- type: model
+  name: JARVIS-1
+  organization: Peking University Institute for Artificial Intelligence
+  description: JARVIS-1 is an open-world agent that can perceive multimodal input (visual observations and human instructions), generate sophisticated plans, and perform embodied control, all within the popular yet challenging open-world Minecraft universe.
+  created_date: 2023-11-10
+  url: https://arxiv.org/pdf/2311.05997.pdf
+  model_card: none
+  modality: text; in-game actions
+  analysis: Compared with other multi-task, instruction-following agents. 
+  size: unknown
+  dependencies: []
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: ''
+  access: open
+  license: unknown
+  intended_uses: ''
+  prohibited_uses: ''
+  monitoring: none
+  feedback: none

assets/tsinghua.yaml

@@ -1,7 +1,7 @@
 ---
 - type: model
   name: CodeGeeX
-  organization: Tsinghua
+  organization: Tsinghua University
   description: CodeGeeX is an autoregressive language model trained on code
   created_date: 2022-09-20
   url: https://github.com/THUDM/CodeGeeX
@@ -26,7 +26,7 @@
   feedback: none
 - type: model
   name: CogView
-  organization: Tsinghua
+  organization: Tsinghua University
   description: CogView is a transformer model for text-to-image generation
   created_date:
     explanation: The date the model paper was released
@@ -53,7 +53,7 @@
   feedback: ''
 - type: model
   name: CogView 2
-  organization: Tsinghua
+  organization: Tsinghua University
   description: CogView 2 is a hierarchical transformer for text-to-image generation
   created_date:
     explanation: The date the model paper was released
@@ -80,7 +80,7 @@
   feedback: ''
 - type: model
   name: CogVideo
-  organization: Tsinghua
+  organization: Tsinghua University
   description: CogVideo is a transformer model for text-to-video generation
   created_date:
     explanation: The date the model paper was released
@@ -107,7 +107,7 @@
   feedback: ''
 - type: model
   name: GLM-130B
-  organization: Tsinghua
+  organization: Tsinghua University
   description: GLM-130B is a bidirectional language model trained on English and
     Chinese
   created_date:
@@ -137,3 +137,25 @@
   prohibited_uses: ''
   monitoring: ''
   feedback: ''
+- type: model
+  name: CogVLM
+  organization: Zhipu AI, Tsinghua University
+  description: CogVLM is a powerful open-source visual language foundation model
+  created_date: 2023-11-06
+  url: https://arxiv.org/pdf/2311.03079.pdf
+  model_card: none
+  modality: image, text; text
+  analysis: Evaluated on image captioning and visual question answering benchmarks.
+  size: 17B parameters (dense)
+  dependencies: [Vicuna, CLIP]
+  training_emissions: unknown
+  training_time: 4096 A100 days
+  training_hardware: unknown
+  quality_control: none
+  access: open
+  license: Apache 2.0
+  intended_uses: Future multimodal research
+  prohibited_uses: none
+  monitoring: none
+  feedback: none
+

assets/xai.yaml

@@ -0,0 +1,22 @@
+---
+- type: model
+  name: Grok-1
+  organization: xAI
+  description: Grok is an AI modeled after the Hitchhiker’s Guide to the Galaxy, intended to answer almost anything and even suggest what questions to ask.
+  created_date: 2023-11-04
+  url: https://x.ai/
+  model_card: https://x.ai/model-card/
+  modality: text; text
+  analysis: Grok-1 was evaluated on a range of reasoning benchmark tasks and on curated foreign mathematic examination questions.
+  size: unknown
+  dependencies: []
+  training_emissions: unknown
+  training_time: unknown
+  training_hardware: unknown
+  quality_control: none
+  access: closed
+  license: unknown
+  intended_uses: Grok-1 is intended to be used as the engine behind Grok for natural language processing tasks including question answering, information retrieval, creative writing and coding assistance.
+  prohibited_uses: none
+  monitoring: unknown
+  feedback: none