Link the arXiv preprint

README.md

@@ -14,6 +14,8 @@ library_name: peft
 Concept-bottleneck controllable-generation modules for **GPT-2 Medium (355M)**, from
 the paper "The Illusion of Control: Why Bare Classifier Inversion Silently Fails in Concept-Bottleneck Text Generation" (EMNLP 2026).
 
+Paper: [arXiv:2608.22956](https://arxiv.org/abs/2608.22956)
+
 Matched-backbone comparison against the CompMCTG baselines (Section 7.1), all of which use GPT-2-Medium. Reported over three training seeds.
 
 This repository contains **only our own trained parameters**: the per-axis
@@ -91,6 +93,10 @@ terms as the corresponding base model.
   author    = {Bing, Qi and Shao, Xiaowei},
   booktitle = {Proceedings of the 2026 Conference on Empirical Methods in
                Natural Language Processing (EMNLP)},
-  year      = {2026}
+  year      = {2026},
+  eprint    = {2608.22956},
+  archivePrefix = {arXiv},
+  primaryClass  = {cs.CL},
+  url       = {https://arxiv.org/abs/2608.22956}
 }
 ```