openbmb/UltraCM-13b
Text Generation
•
Updated
•
29
•
18
Critique Models (CM) on the 🤗 Hub
This collection contains some Critique Models (CM) for LLM evaluation available in the HuggingFace Hub
Note UltraCM was released by OpenBMB as part of their UltraFeedback paper, trained on GPT-4 critiques and scores
prometheus-eval/prometheus-7b-v1.0
Text2Text Generation • Updated • 998 • 28Note Kaist AI released Prometheus, a Critique Model (CM) as an alternative to the usage of GPT-4 from OpenAI as a critique model, which is expensive and may have intra-model bias. This is the fine-tune on top of `meta-llama/Llama-2-7b-chat-hf`, and given an instruction, a response, and a reference answer, generates the `feedback` and `score` for it
prometheus-eval/prometheus-13b-v1.0
Text2Text Generation • Updated • 9.52k • 113Note Kaist AI released Prometheus, a Critique Model (CM) as an alternative to the usage of GPT-4 from OpenAI as a critique model, which is expensive and may have intra-model bias. This is the fine-tune on top of `meta-llama/Llama-2-13b-chat-hf`, and given an instruction, a response, and a reference answer, generates the `feedback` and `score` for it
prometheus-eval/prometheus-7b-v2.0
Text2Text Generation • Updated • 4.86k • 40Note Kaist AI (now under the `prometheus-eval` org) recently released Prometheus 2, this is the `mistralai/Mistral-7B-Instruct-v0.2` fine-tune that generates both `feedback` and `score`, and can run either absolute (single response) or relative (two responses), both with or without a reference answer; as Prometheus 1 was expecting a reference answer in any case
prometheus-eval/prometheus-8x7b-v2.0
Text2Text Generation • Updated • 575 • 32Note Kaist AI (now under the `prometheus-eval` org) recently released Prometheus 2, this is the `mistralai/Mixtral-8x7B-v0.1` fine-tune that generates both `feedback` and `score`, and can run either absolute (single response) or relative (two responses), both with or without a reference answer; as Prometheus 1 was expecting a reference answer in any case