This paper poses two critical issues in evaluating base models (without post-training): (1) Unstable evaluation during training: in the early stages of pre-training, the models lack the capability to answer questions as required, leading to unstable evaluation results. This instability makes it difficult to provide solid conclusions to guide the training, especially for key experiments such as data ablation and scaling law. (2) Inconsistency between base and instruct models: base models generally exhibit poorer evaluation performance compared to corresponding instruct models. This gap poses a challenge for assessing whether a base model with better evaluation can truly lead to a better instruct model. To address these issues, we propose **B**ase model **O**riented **S**ystematic **E**valuation (**BOSE**), a method specifically designed to optimize the evaluation of base models. Specifically, BOSE introduces two key innovations: In-Context Light-instruction Prompt (**ICLiP**) for open-ended tasks and **Blank-ppl** for multi-choice tasks with candidate options, which transforms the standard perplexity (ppl) metric into a fill-in-the-blank format to mitigate early-stage evaluation fluctuations. Furthermore, we are the first to propose Kendall’s rank correlation to quantitatively measure the evaluation stability and consistency. Experimental results demonstrate that BOSE significantly enhances both the stability of evaluations during pre-training and the consistency between base and instruct models, thereby providing more reliable guidance for the LLMs’ training.

The ClimateCheck shared task addresses the critical challenge of grounding social media claims about climate change in scientific literature. This paper details our winning approach. For abstract retrieval, we propose a multi-stage pipeline: (1) initial candidate generation from a corpus of ~400,000 abstracts using BM25; (2) fine-grained reranking of these candidates using an ensemble of BGE-Reranker cross-encoder models, fine-tuned with a specialized training set incorporating both random and hard negative samples; and (3) final list selection based on an RRF-ensembled score. For the verification aspect, we leverage Gemini 2.5 Pro to classify the relationship (Supports, Refutes, Not Enough Information) between claims and the retrieved abstracts, guided by carefully engineered prompts. Our system achieved first place in both subtasks, demonstrating the efficacy of combining robust sparse retrieval, powerful neural rerankers, strategic negative sampling, and LLM-based semantic analysis for connecting social media discourse to scientific evidence. Part of the example code: https://anonymous.4open.science/r/climatecheck_solution-1120

Chinese Spelling Check (CSC) is a task to detect and correct spelling errors in Chinese natural language. Existing methods have made attempts to incorporate the similarity knowledge between Chinese characters. However, they take the similarity knowledge as either an external input resource or just heuristic rules. This paper proposes to incorporate phonological and visual similarity knowledge into language models for CSC via a specialized graph convolutional network (SpellGCN). The model builds a graph over the characters, and SpellGCN is learned to map this graph into a set of inter-dependent character classifiers. These classifiers are applied to the representations extracted by another network, such as BERT, enabling the whole network to be end-to-end trainable. Experiments are conducted on three human-annotated datasets. Our method achieves superior performance against previous models by a large margin.

The ambiguous annotation criteria lead to divergence of Chinese Word Segmentation (CWS) datasets in various granularities. Multi-criteria Chinese word segmentation aims to capture various annotation criteria among datasets and leverage their common underlying knowledge. In this paper, we propose a domain adaptive segmenter to exploit diverse criteria of various datasets. Our model is based on Bidirectional Encoder Representations from Transformers (BERT), which is responsible for introducing open-domain knowledge. Private and shared projection layers are proposed to capture domain-specific knowledge and common knowledge, respectively. We also optimize computational efficiency via distillation, quantization, and compiler optimization. Experiments show that our segmenter outperforms the previous state of the art (SOTA) models on 10 CWS datasets with superior efficiency.

Numerical reasoning over texts, such as addition, subtraction, sorting and counting, is a challenging machine reading comprehension task, since it requires both natural language understanding and arithmetic computation. To address this challenge, we propose a heterogeneous graph representation for the context of the passage and question needed for such reasoning, and design a question directed graph attention network to drive multi-step numerical reasoning over this context graph. Our model, which combines deep learning and graph reasoning, achieves remarkable results in benchmark datasets such as DROP.

Aspect-term sentiment analysis (ATSA) is a long-standing challenge in natural language process. It requires fine-grained semantical reasoning about a target entity appeared in the text. As manual annotation over the aspects is laborious and time-consuming, the amount of labeled data is limited for supervised learning. This paper proposes a semi-supervised method for the ATSA problem by using the Variational Autoencoder based on Transformer. The model learns the latent distribution via variational inference. By disentangling the latent representation into the aspect-specific sentiment and the lexical context, our method induces the underlying sentiment prediction for the unlabeled data, which then benefits the ATSA classifier. Our method is classifier-agnostic, i.e., the classifier is an independent module and various supervised models can be integrated. Experimental results are obtained on the SemEval 2014 task 4 and show that our method is effective with different the five specific classifiers and outperforms these models by a significant margin.