Human experts tackle difficult math problems by identifying and executing a few pivotal steps rather than listing every intermediate thought. In contrast, standard Chain-of-Thought (CoT) distillation trains small models on lengthy reasoning traces, encouraging a uniform overthinking style across easy and hard items alike. The result is rigid, slow solutions that sacrifice adaptivity. This approach stands in sharp contrast to human intuition. Humans naturally adapt their problem-solving strategy, dedicating significant effort to difficult problems while finding quick, simple solutions for easier ones. We argue that the root cause lies in the training data: it contains excess information and reasoning steps organized in ways misaligned with human practice. We address this with Difficulty-Aware Distillation(DAD), a procedure for producing training data that mirrors concise human reasoning. A large teacher model first assesses a problem’s difficulty and then rewrites the solution to retain only the essential steps. Using this process, we constructed LiteCoT, a 100,000-example corpus of short, clear rationales, and used it to train our Liter models. With 100k LiteCoT, we outperform models trained on 800k long CoT and cut both training and inference costs. The advantage is consistent across standard math benchmarks, showing that concise, human-aligned data delivers equal or better accuracy with much less compute. For example, on the challenging AIME24 exam, our approach reaches 74.2% Pass@1 using only about 5K inference tokens, surpassing other methods that consume many more tokens.

In this paper, we report the solution of the team BERT 4EVER for the LT-EDI-2022 shared task2: Homophobia/Transphobia Detection in social media comments in ACL 2022, which aims to classify Youtube comments into one of the following categories: no,moderate, or severe depression. We model the problem as a text classification task and a text generation task and respectively propose two different models for the tasks. To combine the knowledge learned from these two different models, we softly fuse the predicted probabilities of the models above and then select the label with the highest probability as the final output. In addition, multiple augmentation strategies are leveraged to improve the model generalization capability, such as back translation and adversarial training. Experimental results demonstrate the effectiveness of the proposed models and two augmented strategies.