Going multimodal: How Prezi is leveraging the Hub and the Expert Support Program to accelerate their ML roadmap
β’
11
Moritz Laurer
MoritzLaurer
AI & ML interests
None yet
Recent Activity
liked
a model
about 6 hours ago
hexgrad/Kokoro-82M
new activity
about 6 hours ago
Pendrokar/TTS-Spaces-Arena:What to vote when quality of generation is equal or both are bad?
Articles
Organizations
Posts
14
Post
1131
FACTS is a great paper from
@GoogleDeepMind
on measuring the factuality of LLM outputs. You can now download their prompt templates from
@huggingface
to improve LLM-based fact-checking yourself!
π The paper introduces the FACTS Grounding benchmark for evaluating the factuality of LLM outputs.
π€ Fact-checking is automated by an ensemble of LLM judges that verify if a response is fully grounded in a factual reference document.
π§ͺ The authors tested different prompt templates on held-out data to ensure their generalization.
π It's highly educational to read these templates to learn how frontier labs design prompts and understand their limitations.
πΎ You can now download and reuse these prompt templates via the prompt-templates library!
π The library simplifies sharing prompt templates on the HF hub or locally via standardized YAML files. Letβs make LLM work more transparent and reproducible by sharing more templates like this!
Links π
- prompt-templates docs: https://moritzlaurer.github.io/prompt_templates/
- all templates on the HF Hub: MoritzLaurer/facts-grounding-prompts
- FACTS paper: https://storage.googleapis.com/deepmind-media/FACTS/FACTS_grounding_paper.pdf
π The paper introduces the FACTS Grounding benchmark for evaluating the factuality of LLM outputs.
π€ Fact-checking is automated by an ensemble of LLM judges that verify if a response is fully grounded in a factual reference document.
π§ͺ The authors tested different prompt templates on held-out data to ensure their generalization.
π It's highly educational to read these templates to learn how frontier labs design prompts and understand their limitations.
πΎ You can now download and reuse these prompt templates via the prompt-templates library!
π The library simplifies sharing prompt templates on the HF hub or locally via standardized YAML files. Letβs make LLM work more transparent and reproducible by sharing more templates like this!
Links π
- prompt-templates docs: https://moritzlaurer.github.io/prompt_templates/
- all templates on the HF Hub: MoritzLaurer/facts-grounding-prompts
- FACTS paper: https://storage.googleapis.com/deepmind-media/FACTS/FACTS_grounding_paper.pdf
Post
1531
The TRL v0.13 release is π₯! My highlight are the new process reward trainer to train models similar to o1 and tool call support:
π§ Process reward trainer: Enables training of Process-supervised Reward Models (PRMs), which reward the quality of intermediate steps, promoting structured reasoning. Perfect for tasks like stepwise reasoning.
π Model merging: A new callback leverages mergekit to merge models during training, improving performance by blending reference and policy models - optionally pushing merged models to the Hugging Face Hub.
π οΈ Tool call support: TRL preprocessing now supports tool integration, laying the groundwork for agent fine-tuning with examples like dynamic temperature fetching in prompts.
βοΈ Mixture of judges: The new AllTrueJudge combines decisions from multiple binary judges for more nuanced evaluation.
Read the release notes and other resources here π
Release: https://github.com/huggingface/trl/releases/tag/v0.13.0
Mergekit: https://github.com/arcee-ai/mergekit
Mixture of judges paper: The Perfect Blend: Redefining RLHF with Mixture of Judges (2409.20370)
π§ Process reward trainer: Enables training of Process-supervised Reward Models (PRMs), which reward the quality of intermediate steps, promoting structured reasoning. Perfect for tasks like stepwise reasoning.
π Model merging: A new callback leverages mergekit to merge models during training, improving performance by blending reference and policy models - optionally pushing merged models to the Hugging Face Hub.
π οΈ Tool call support: TRL preprocessing now supports tool integration, laying the groundwork for agent fine-tuning with examples like dynamic temperature fetching in prompts.
βοΈ Mixture of judges: The new AllTrueJudge combines decisions from multiple binary judges for more nuanced evaluation.
Read the release notes and other resources here π
Release: https://github.com/huggingface/trl/releases/tag/v0.13.0
Mergekit: https://github.com/arcee-ai/mergekit
Mixture of judges paper: The Perfect Blend: Redefining RLHF with Mixture of Judges (2409.20370)
Collections
7
Repos with prompts that can be loaded with the `prompt-templates` library. Docs: https://github.com/MoritzLaurer/prompt_templates
These are my current best zeroshot classifiers. Some of my older models are downloaded more often, but the models in this collection are newer/better.
-
MoritzLaurer/ModernBERT-large-zeroshot-v2.0
Text Classification β’ Updated β’ 1.42k β’ 23 -
MoritzLaurer/deberta-v3-large-zeroshot-v2.0
Zero-Shot Classification β’ Updated β’ 55.5k β’ 87 -
MoritzLaurer/bge-m3-zeroshot-v2.0
Zero-Shot Classification β’ Updated β’ 1.5M β’ β’ 43 -
MoritzLaurer/deberta-v3-base-zeroshot-v2.0
Zero-Shot Classification β’ Updated β’ 1.79k β’ 8
Papers
1
spaces
10
models
42
MoritzLaurer/ModernBERT-base-zeroshot-v2.0
Text Classification
β’
Updated
β’
291
β’
12
MoritzLaurer/ModernBERT-large-zeroshot-v2.0
Text Classification
β’
Updated
β’
1.42k
β’
23
MoritzLaurer/open_models_special_prompts
Updated
MoritzLaurer/parler-tts-large-v1
Text-to-Speech
β’
Updated
β’
15
MoritzLaurer/parler-tts-mini-v1.1
Text-to-Speech
β’
Updated
β’
5
MoritzLaurer/deberta-v3-base-zeroshot-v1.1-all-33
Zero-Shot Classification
β’
Updated
β’
448k
β’
26
MoritzLaurer/deberta-v3-xsmall-zeroshot-v1.1-all-33
Zero-Shot Classification
β’
Updated
β’
28.8k
β’
4
MoritzLaurer/bge-m3-zeroshot-v2.0
Zero-Shot Classification
β’
Updated
β’
1.5M
β’
β’
43
MoritzLaurer/xlm-v-base-mnli-xnli
Zero-Shot Classification
β’
Updated
β’
36
β’
20
MoritzLaurer/multilingual-MiniLMv2-L6-mnli-xnli
Zero-Shot Classification
β’
Updated
β’
275k
β’
37
datasets
29
MoritzLaurer/rstar-math-prompts
Updated
MoritzLaurer/distiset-test
Viewer
β’
Updated
β’
50
β’
14
MoritzLaurer/facts-grounding-prompts
Updated
β’
49
β’
2
MoritzLaurer/example_prompts
Updated
β’
142
MoritzLaurer/prompts_from_papers
Updated
β’
80
MoritzLaurer/closed_system_prompts
Updated
β’
148
MoritzLaurer/dataset_prompts
Updated
β’
44
MoritzLaurer/example_prompts_test
Updated
β’
104
MoritzLaurer/example_tools
Updated
β’
63
MoritzLaurer/upload-test
Viewer
β’
Updated
β’
6
β’
34