anonymous update

README.md

@@ -6,8 +6,7 @@ Activation_Decoding: Decoding by sharpness inside Large Language Models
 
 ![](figure/ad.png)
 
-This is the code implementation of the paper: [In-Context Sharpness as Alerts: An Inner Representation Perspective for Hallucination Mitigation
-](https://arxiv.org/abs/2403.01548)
+This is the code implementation of the paper: In-Context Sharpness as Alerts: An Inner Representation Perspective for Hallucination Mitigation
 
 We discover a pattern associated with hallucinations: correct generations tend to have *sharper* context activations in the hidden states of the in-context tokens, compared to that of the incorrect generations. Leveraging this signal, we propose an entropy-based metric to quantify the "*sharpness* " among the in-context hidden states and incorporate it into the decoding process, i.e, use the entropy value to adjust the next token prediction distribution to improve the factuality and overall quality of the generated text.
 
@@ -38,15 +37,6 @@ pip install -r requirements.txt
 | `--num-gpus`      | `1` | Number of GPUs to use, `1/1/2` for `7B/13B/70B` model sizes respectively if you are using 80G GPU cards.  |
 | `--do-rating`      | | Add this to output evaluation results  |
 
-#### In-domain Hyperparameter choice
-
-In the table below, we present the `alpha` with the `info_layer` choices for each knowledge-seeking dataset, formatted as `alpha/info_layer`.
-
-| | TriviaQA | HotpotQA | NQ |
-|---| --- | --- | --- |
-|`LLaMa2-7B-chat`| `1/32`| `0.8/32` |`0.2/32`  |
-|`LLaMa2-13B-chat`| `0.6/32` |`1/28`  |  `0.8/34`|
-|`LLaMa2-70B-chat`|  `1/70`| `0.6/66` | `1/68` |
 
 ### TriviaQA, HotpotQA and Natural Questions
 
@@ -94,20 +84,4 @@ bash run_tqa.sh # refer to this script for more details
 
 - Dola: https://github.com/voidism/DoLa
 
-## Citation
-
-Please cite our paper if it's helpful to your work!
-
-```bibtex
-@inproceedings{
-chen2024incontext,
-title={In-Context Sharpness as Alerts: An Inner Representation Perspective for Hallucination Mitigation},
-author={Chen, Shiqi and Xiong, Miao and Liu, Junteng and Wu, Zhengxuan and Xiao, Teng and Gao, Siyang and He, Junxian},
-booktitle={Arxiv},
-year={2024},
-url={https://arxiv.org/abs/2403.01548}
-}
-
-```
-