Peer-review-in-LLMs: Automatic Evaluation Method for LLMs in Open-environment

Existing large language models (LLMs) evaluation methods typically focus on testing the performance on some closed-environment and domain-specific benchmarks with human annotations. In this paper, we explore a novel unsupervised evaluation direction, utilizing peer-review mechanisms to measure LLMs automatically. In this setting, both open-source and closed-source LLMs lie in the same environment, capable of answering unlabeled questions and evaluating each other, where each LLM's response score is jointly determined by other anonymous ones. To obtain the ability hierarchy among these models, we assign each LLM a learnable capability parameter to adjust the final ranking. We formalize it as a constrained optimization problem, intending to maximize the consistency of each LLM's capabilities and scores. The key assumption behind is that high-level LLM can evaluate others' answers more accurately than low-level ones, while higher-level LLM can also achieve higher response scores. Moreover, we propose three metrics called PEN, CIN, and LIS to evaluate the gap in aligning human rankings. We perform experiments on multiple datasets with these metrics, validating the effectiveness of the proposed approach.

Towards Better Query Classification with Multi-Expert Knowledge Condensation in JD Ads Search

Search query classification, as an effective way to understand user intents, is of great importance in real-world online ads systems. To ensure a lower latency, a shallow model (e.g. FastText) is widely used for efficient online inference. However, the representation ability of the FastText model is insufficient, resulting in poor classification performance, especially on some low-frequency queries and tailed categories. Using a deeper and more complex model (e.g. BERT) is an effective solution, but it will cause a higher online inference latency and more expensive computing costs. Thus, how to juggle both inference efficiency and classification performance is obviously of great practical importance. To overcome this challenge, in this paper, we propose knowledge condensation (KC), a simple yet effective knowledge distillation framework to boost the classification performance of the online FastText model under strict low latency constraints. Specifically, we propose to train an offline BERT model to retrieve more potentially relevant data. Benefiting from its powerful semantic representation, more relevant labels not exposed in the historical data will be added into the training set for better FastText model training. Moreover, a novel distribution-diverse multi-expert learning strategy is proposed to further improve the mining ability of relevant data. By training multiple BERT models from different data distributions, it can respectively perform better at high, middle, and low-frequency search queries. The model ensemble from multi-distribution makes its retrieval ability more powerful. We have deployed two versions of this framework in JD search, and both offline experiments and online A/B testing from multiple datasets have validated the effectiveness of the proposed approach.

Data Provenance via Differential Auditing

Auditing Data Provenance (ADP), i.e., auditing if a certain piece of data has been used to train a machine learning model, is an important problem in data provenance. The feasibility of the task has been demonstrated by existing auditing techniques, e.g., shadow auditing methods, under certain conditions such as the availability of label information and the knowledge of training protocols for the target model. Unfortunately, both of these conditions are often unavailable in real applications. In this paper, we introduce Data Provenance via Differential Auditing (DPDA), a practical framework for auditing data provenance with a different approach based on statistically significant differentials, i.e., after carefully designed transformation, perturbed input data from the target model's training set would result in much more drastic changes in the output than those from the model's non-training set. This framework allows auditors to distinguish training data from non-training ones without the need of training any shadow models with the help of labeled output data. Furthermore, we propose two effective auditing function implementations, an additive one and a multiplicative one. We report evaluations on real-world data sets demonstrating the effectiveness of our proposed auditing technique.

Unorganized Malicious Attacks Detection

Recommender system has attracted much attention during the past decade. Many attack detection algorithms have been developed for better recommendations, mostly focusing on shilling attacks, where an attack organizer produces a large number of user profiles by the same strategy to promote or demote an item. This work considers a different attack style: unorganized malicious attacks, where attackers individually utilize a small number of user profiles to attack different items without any organizer. This attack style occurs in many real applications, yet relevant study remains open. We first formulate the unorganized malicious attacks detection as a matrix completion problem, and propose the Unorganized Malicious Attacks detection (UMA) approach, a proximal alternating splitting augmented Lagrangian method. We verify, both theoretically and empirically, the effectiveness of our proposed approach.