Evaluating Large Language Models (LLMs) has become increasingly important, with automatic evaluation benchmarks gaining prominence as alternatives to human evaluation. While existing research has focused on approximating model rankings, such benchmarks fail to provide users and developers with a comprehensive and fine-grained understanding of a specific model’s capabilities. To fill this gap, we propose SCAN (Structured Capability Assessment and Navigation), a practical framework that enables detailed characterization of LLM capabilities through comprehensive and fine-grained evaluation. SCAN incorporates four key components: (1) TaxBuilder, which extracts capability-indicating tags from extensive queries to construct a hierarchical taxonomy automatically; (2) RealMix, a query synthesis and filtering mechanism that ensures sufficient evaluation data for each capability tag; (3) a suite of visualization and analysis tools that facilitate efficient navigation and analysis of model capabilities; and (4) a PC²-based (Pre-Comparison-derived Criteria) LLM-as-a-Judge approach that achieves significantly higher accuracy compared to classic LLM-as-a-Judge method. Using SCAN, we conduct a comprehensive evaluation of 21 mainstream LLMs. Our detailed analysis of the GPT-OSS family reveals substantial performance variations, even within sub-capabilities belonging to the same category of capability. This finding highlights the importance of fine-grained evaluation in accurately understanding LLM behavior. Project homepage and resources are available at https://github.com/liudan193/SCAN.

In this paper, we propose an unsupervised query enhanced approach for knowledge-intensive conversations, namely QKConv. There are three modules in QKConv: a query generator, an off-the-shelf knowledge selector, and a response generator. QKConv is optimized through joint training, which produces the response by exploring multiple candidate queries and leveraging corresponding selected knowledge. The joint training solely relies on the dialogue context and target response, getting exempt from extra query annotations or knowledge provenances. To evaluate the effectiveness of the proposed QKConv, we conduct experiments on three representative knowledge-intensive conversation datasets: conversational question-answering, task-oriented dialogue, and knowledge-grounded conversation. Experimental results reveal that QKConv performs better than all unsupervised methods across three datasets and achieves competitive performance compared to supervised methods.

Existing pipelined task-oriented dialogue systems usually have difficulties adapting to unseen domains, whereas end-to-end systems are plagued by large-scale knowledge bases in practice. In this paper, we introduce a novel query-driven task-oriented dialogue system, namely Q-TOD. The essential information from the dialogue context is extracted into a query, which is further employed to retrieve relevant knowledge records for response generation. Firstly, as the query is in the form of natural language and not confined to the schema of the knowledge base, the issue of domain adaption is alleviated remarkably in Q-TOD. Secondly, as the query enables the decoupling of knowledge retrieval from the generation, Q-TOD gets rid of the issue of knowledge base scalability. To evaluate the effectiveness of the proposed Q-TOD, we collect query annotations for three publicly available task-oriented dialogue datasets. Comprehensive experiments verify that Q-TOD outperforms strong baselines and establishes a new state-of-the-art performance on these datasets.

To explore the limit of dialogue generation pre-training, we present the models of PLATO-XL with up to 11 billion parameters, trained on both Chinese and English social media conversations. To train such large models, we adopt the architecture of unified transformer with high computation and parameter efficiency. In addition, we carry out multi-party aware pre-training to better distinguish the characteristic information in social media conversations. With such designs, PLATO-XL successfully achieves superior performances as compared to other approaches in both Chinese and English chitchat. We further explore the capacity of PLATO-XL on other conversational tasks, such as knowledge grounded dialogue and task-oriented conversation. The experimental results indicate that PLATO-XL obtains state-of-the-art results across multiple conversational tasks, verifying its potential as a foundation model of conversational AI.

Pre-trained language models (PLMs) can provide a good starting point for downstream applications. However, it is difficult to generalize PLMs to new tasks given a few labeled samples. In this work, we show that Relation Graph augmented Learning (RGL) can improve the performance of few-shot natural language understanding tasks. During learning, RGL constructs a relation graph based on the label consistency between samples in the same batch, and learns to solve the resultant node classification and link prediction problems on the relation graph. In this way, RGL fully exploits the limited supervised information, which can boost the tuning effectiveness. Extensive experimental results show that RGL consistently improves the performance of prompt-based tuning strategies.

In task-oriented dialogue systems, recent dialogue state tracking methods tend to perform one-pass generation of the dialogue state based on the previous dialogue state. The mistakes of these models made at the current turn are prone to be carried over to the next turn, causing error propagation. In this paper, we propose a novel Amendable Generation for Dialogue State Tracking (AG-DST), which contains a two-pass generation process: (1) generating a primitive dialogue state based on the dialogue of the current turn and the previous dialogue state, and (2) amending the primitive dialogue state from the first pass. With the additional amending generation pass, our model is tasked to learn more robust dialogue state tracking by amending the errors that still exist in the primitive dialogue state, which plays the role of reviser in the double-checking process and alleviates unnecessary error propagation. Experimental results show that AG-DST significantly outperforms previous works in two active DST datasets (MultiWOZ 2.2 and WOZ 2.0), achieving new state-of-the-art performances.