gsarti commited on
Commit
e512909
•
1 Parent(s): 88caca3

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +11 -3
README.md CHANGED
@@ -38,9 +38,9 @@ model-index:
38
  name: Solution Exact Match
39
  ---
40
 
41
- # Phi-3 Mini 4K Verbalized Rebus Solver 🇮🇹
42
 
43
- This model is a parameter-efficient fine-tuned version of Phi-3 Mini 4K trained for verbalized rebus solving in Italian, as part of the [release](https://huggingface.co/collections/gsarti/verbalized-rebus-clic-it-2024-66ab8f11cb04e68bdf4fb028) for our paper [Non Verbis, Sed Rebus: Large Language Models are Weak Solvers of Italian Rebuses](TBD). The task of verbalized rebus solving consists of converting an encrypted sequence of letters and crossword definitions into a solution phrase matching the word lengths specified in the solution key. An example is provided below.
44
 
45
  The model was trained in 4-bit precision for 5070 steps on the verbalized subset of the [EurekaRebus](https://huggingface.co/datasets/gsarti/eureka-rebus) using QLora via [Unsloth](https://github.com/unslothai/unsloth) and [TRL](https://github.com/huggingface/trl). This repository contains PEFT-compatible adapters saved throughout training. Use the `revision=<GIT_HASH>` parameter in `from_pretrained` to load mid-training adapter checkpoints.
46
 
@@ -127,7 +127,15 @@ For problems or updates on this model, please contact [[email protected]
127
  If you use this model in your work, please cite our paper as follows:
128
 
129
  ```bibtex
130
- TBD
 
 
 
 
 
 
 
 
131
  ```
132
 
133
  ## Acknowledgements
 
38
  name: Solution Exact Match
39
  ---
40
 
41
+ # Phi-3 Mini 4K Verbalized Rebus Solver - PEFT Adapters 🇮🇹
42
 
43
+ This model is a parameter-efficient fine-tuned version of Phi-3 Mini 4K trained for verbalized rebus solving in Italian, as part of the [release](https://huggingface.co/collections/gsarti/verbalized-rebus-clic-it-2024-66ab8f11cb04e68bdf4fb028) for our paper [Non Verbis, Sed Rebus: Large Language Models are Weak Solvers of Italian Rebuses](https://arxiv.org/abs/2408.00584). The task of verbalized rebus solving consists of converting an encrypted sequence of letters and crossword definitions into a solution phrase matching the word lengths specified in the solution key. An example is provided below.
44
 
45
  The model was trained in 4-bit precision for 5070 steps on the verbalized subset of the [EurekaRebus](https://huggingface.co/datasets/gsarti/eureka-rebus) using QLora via [Unsloth](https://github.com/unslothai/unsloth) and [TRL](https://github.com/huggingface/trl). This repository contains PEFT-compatible adapters saved throughout training. Use the `revision=<GIT_HASH>` parameter in `from_pretrained` to load mid-training adapter checkpoints.
46
 
 
127
  If you use this model in your work, please cite our paper as follows:
128
 
129
  ```bibtex
130
+ @article{sarti-etal-2024-rebus,
131
+ title = "Non Verbis, Sed Rebus: Large Language Models are Weak Solvers of Italian Rebuses",
132
+ author = "Sarti, Gabriele and Caselli, Tommaso and Nissim, Malvina and Bisazza, Arianna",
133
+ journal = "ArXiv",
134
+ month = jul,
135
+ year = "2024",
136
+ volume = {abs/2408.00584},
137
+ url = {https://arxiv.org/abs/2408.00584},
138
+ }
139
  ```
140
 
141
  ## Acknowledgements