Canada

Data contamination in evaluation is getting increasingly prevalent with the emerge of language models pre-trained on super large, automatically-crawled corpora. This problem leads to significant challenges in accurate assessment of model capabilities and generalisations. In this paper, we propose LatestEval, an automatic method leverages the most recent texts to create uncontaminated reading comprehension evaluations. LatestEval avoids data contamination by only using texts published within a recent time window, ensuring no overlap with the training corpora of pre-trained language models. We develop LatestEval automated pipeline to 1) gather latest texts; 2) identify key information, and 3) construct questions targeting the information while removing the existing answers from the context. This encourages models to infer the answers themselves based on the remaining context, rather than just copy-paste. Our experiments demonstrate that language models exhibit negligible memorisation behaviours on LatestEval as opposed to previous benchmarks, suggesting a significantly reduced risk of data contamination.

AAAI 2024

LatestEval: Addressing Data Contamination in Language Model Evaluation through Dynamic and Time-Sensitive Test Construction

large language models

evaluation

### Welcome!
The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. Our Virtual Poster Sessions will take place online Thursday, August 22, 2024.

You are required to register for this event. **Please register [here](https://2024.aclweb.org/registration). **

If you have already registered, please check your inbox for an email from Underline granting you access to ACL 2024 content.

Please register!

ACL 2024

The 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) will take place in Bangkok, Thailand from August 11th to 16th, 2024. More information will be announced soon.

technical paper

We are pleased to announce the Thirty-Eighth AAAI Conference on Artificial Intelligence (AAAI-24), which will be held in Vancouver, British Columbia at the Vancouver Convention Centre – West Building from 20-27 February, 2024.

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-24 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and a range of other activities to be announced.

We expect for AAAI-24 to be an in-person conference – one author of all accepted papers will be expected to present work in person unless there are exceptional circumstances that prevent this.

In order to access the AAAI-24 event page you need to register [here](https://aaai.org/aaai-conference/registration/)

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. 

Crowdsourcing platforms use various truth discovery algorithms to aggregate annotations from multiple labelers. In an online setting, however, the main challenge is to decide whether to ask for more annotations for each item to efficiently trade off cost (i.e., the number of annotations) for quality of the aggregated annotations. In this paper, we propose a novel approach for general complex annotation (such as bounding boxes and taxonomy paths), that works in an online crowdsourcing setting. We prove that the expected average similarity of a labeler is linear in their accuracy conditional on the reported label. This enables us to infer reported label accuracy in a broad range of scenarios. We conduct extensive evaluations on real-world crowdsourcing data from Meta and show the effectiveness of our proposed online algorithms in improving the cost-quality trade-off.

Efficient Online Crowdsourcing with Complex Annotations

This paper develops a novel rating-based reinforcement learning approach that uses human ratings to obtain human guidance in reinforcement learning. Different from the existing preference-based and ranking-based reinforcement learning paradigms, based on human relative preferences over sample pairs, the proposed rating-based reinforcement learning approach is based on human evaluation of individual trajectories without relative comparisons between sample pairs. The rating-based reinforcement learning approach builds on a new prediction model for human ratings and a novel multi-class loss function. We conduct several experimental studies based on synthetic ratings and real human ratings to evaluate the effectiveness and benefits of the new rating-based reinforcement learning approach.

Rating-Based Reinforcement Learning

Artificial intelligence (AI) has made remarkable progress across various domains, with large language models like ChatGPT gaining substantial attention for their human-like text-generation capabilities. Despite these achievements, improving spatial reasoning remains a significant challenge for these models. Benchmarks like StepGame evaluate AI spatial reasoning, where ChatGPT has shown unsatisfactory performance. However, the presence of template errors in the benchmark has an impact on the evaluation results. Thus there is potential for ChatGPT to perform better if these template errors are addressed, leading to more accurate assessments of its spatial reasoning capabilities. In this study, we refine the StepGame benchmark, providing a more accurate dataset for model evaluation. We analyze GPT’s spatial reasoning performance on the rectified benchmark, identifying proficiency in mapping natural language text to spatial relations but limitations in multi-hop reasoning. We provide a flawless solution to the benchmark by combining template-to-relation mapping with logic-based reasoning. This combination demonstrates proficiency in performing qualitative reasoning on StepGame without encountering any errors. We then address the limitations of GPT models in spatial reasoning. To improve spatial reasoning, we deploy Chain-of-Thought and Tree-of-thoughts prompting strategies, offering insights into GPT’s cognitive process. Our investigation not only sheds light on model deficiencies but also proposes enhancements, contributing to the advancement of AI with more robust spatial reasoning capabilities.

Advancing Spatial Reasoning in Large Language Models: An In-Depth Evaluation and Enhancement Using the StepGame Benchmark

Task-oriented dialog systems have witnessed substantial progress due to conversational pre-training techniques. Yet, two significant challenges persist. First, most systems primarily utilize the latest turn’s state label for the generator. This practice overlooks the comprehensive value of state labels in boosting the model’s understanding for future generations. Second, an overreliance on generated policy often leads to error accumulation, resulting in suboptimal responses when adhering to incorrect actions. To combat these challenges, we propose turn-level multi-task objectives for the encoder. With the guidance of essential information from labeled intermediate states, we establish a more robust representation for both understanding and generation. For the decoder, we introduce
an action tree-based scheduled sampling technique. Specifically, inspired by SPACE (He et al. 2022a), we model the hierarchical policy as trees and utilize the similarity between trees to sample negative policy based on scheduled sampling, hoping the model to generate invariant responses under perturbations. This method simulates potential pitfalls by sampling similar negative policy, bridging the gap between task-oriented dialog training and inference. Among methods without continual pre-training, our approach achieved state-of-the-art (SOTA) performance on the MultiWOZ dataset series and was also competitive with pre-trained SOTA methods.

TA&AT: Enhancing Task-Oriented Dialog with Turn-Level Auxiliary Tasks and Action-Tree Based Scheduled Sampling

Satellite data bolstered by their increasing accessibility is leading to many endeavors of automated monitoring of the earth's surface for various applications. Such applications demand high spatial resolution images at a temporal resolution of a few days which entails the challenge of processing a huge volume of image time series data. To overcome this computing bottleneck, we present PatchNet, a bespoke adaptation of beam search and attention mechanism. PatchNet is an automated patch selection neural network that requires only a partial spatial traversal of an image time series and yet achieves impressive results. Satellite systems face a trade-off between spatial and temporal resolutions due to budget/technical constraints e.g., Landsat-8/9 or Sentinel-2 have high spatial resolution whereas, MODIS has high temporal resolution. To deal with the limitation of coarse temporal resolution, we propose FuSITSNet, a twofold feature-based generic fusion model with multimodal learning in a contrastive setting. It produces a learned representation after fusion of two satellite image time series leveraging finer spatial resolution of Landsat and finer temporal resolution of MODIS. The patch alignment module of FuSITSNet aligns the PatchNet processed patches of Landsat-8 with the corresponding MODIS regions to incorporate its finer resolution temporal features. The untraversed patches are handled by the cross-modality attention which highlights additional hot spot features from the two modalities. We conduct extensive experiments on more than 2000 counties of US for crop yield, snow cover, and solar energy prediction and show that even one-fourth spatial processing of image time series produces state-of-the-art results.  FuSITSNet outperforms the predictions of single modality and data obtained using existing generative fusion models and allows for monitoring of dynamic phenomena using freely accessible images, thereby unlocking new opportunities.

Efficient Representation Learning of Satellite Image Time Series and Their Fusion for Spatiotemporal Applications

Multi-relational clustering is a challenging task due to the fact that diverse semantic information conveyed in multi-layer graphs is difficult to extract and fuse. Recent methods integrate topology structure and node attribute information through graph filtering. However, they often use a low-pass filter without fully considering the correlation among multiple graphs. To overcome this drawback, we propose to learn a graph filter motivated by the theoretical analysis of Barlow Twins. We find that input with a negative semi-definite inner product provides a lower bound for Barlow Twins loss, which prevents it from reaching a better solution. We thus learn a filter that yields an upper bound for Barlow Twins. Afterward, we design a simple clustering architecture and demonstrate its state-of-the-art performance on four benchmark datasets. The source code is available at https://github.com/XweiQ/BTGF.

Upper Bounding Barlow Twins: A Novel Filter for Multi-Relational Clustering

In spite of the rapidly evolving landscape of text-to-image generation, the synthesis and manipulation of multiple entities while adhering to specific relational constraints pose enduring challenges. This paper introduces an innovative progressive synthesis and editing operation that systematically incorporates entities into the target image, ensuring their adherence to spatial and relational constraints at each sequential step. Our key insight stems from the observation that while a pre-trained text-to-image diffusion model adeptly handles one or two entities, it often falters when dealing with a greater number. To address this limitation, we propose harnessing the capabilities of a Large Language Model (LLM) to decompose intricate and protracted text descriptions into coherent directives adhering to stringent formats. To facilitate the execution of directives involving distinct semantic operations—namely insertion, editing, and erasing—we formulate the Stimulus, Response, and Fusion (SRF) framework. Within this framework, latent regions are gently stimulated in alignment with each operation, followed by the fusion of the responsive latent components to achieve cohesive entity manipulation. Our proposed framework yields notable advancements in object synthesis, particularly when confronted with intricate and lengthy textual inputs. Consequently, it establishes a new benchmark for text-to-image generation tasks, further elevating the field's performance standards.

Progressive Text-to-Image Diffusion with Soft Latent Direction

In this paper, the worst-case probability measure over the data is introduced as a tool for characterizing the generalization capabilities of machine learning algorithms. More specifically, the worst-case probability measure is a Gibbs probability measure and the unique solution to the maximization of the expected loss under a relative entropy constraint with respect to a reference probability measure. Fundamental generalization metrics, such as the sensitivity of the expected loss, the sensitivity of the empirical risk, and the generalization gap are shown to have closed-form expressions involving the worst-case data-generating probability measure. Existing results for the Gibbs algorithm, such as characterizing the generalization gap as a sum of mutual information and lautum information, up to a constant factor, are recovered. A novel parallel is established between the worst-case data-generating probability measure and the Gibbs algorithm. Specifically, the Gibbs probability measure is identified as a fundamental commonality of the model space and the data space for machine learning algorithms.

Generalization Analysis of Machine Learning Algorithms via the Worst-Case Data-Generating Probability Measure | VIDEO

We study reinforcement learning (RL) in episodic MDPs with adversarial full-information losses and the unknown transition. Instead of the classical static regret, we adopt \emph{dynamic regret} as the performance measure which benchmarks the learner's performance with \emph{changing} policies, making it more suitable for non-stationary environments. The primary challenge is to handle the uncertainties of unknown transition and unknown non-stationarity of environments simultaneously. We propose a general framework to decouple the two sources of uncertainties and show the dynamic regret bound naturally decomposes into two terms, one due to constructing confidence sets to handle the unknown transition and the other due to choosing sub-optimal policies under the unknown non-stationarity. To this end, we first employ the two-layer online ensemble structure to handle the adaptation error due to the unknown non-stationarity, which is model-agnostic. Subsequently, we instantiate the framework to three fundamental MDP models, including tabular MDPs, linear MDPs and linear mixture MDPs, and present corresponding approaches to control the exploration error due to the unknown transition. We provide dynamic regret guarantees respectively and show they are optimal in terms of the number of episodes $K$ and the non-stationarity $\bar{P}_K$ by establishing matching lower bounds. To the best of our knowledge, this is the first work that achieves the optimal (w.r.t. $K$ and $\bar{P}_K$) dynamic regret \emph{without} prior knowledge about the non-stationarity of environments for adversarial MDPs with the unknown transition.

Dynamic Regret of Adversarial MDPs with Unknown Transition and Linear Function Approximation

Recently, there has been growing interest in using Large Language Models (LLMs) for scientific research. Numerous benchmarks have been proposed to evaluate the ability of LLMs for scientific research. However, current benchmarks are mostly based on pre-collected objective questions. This design suffers from data leakage problem and lacks the evaluation of subjective Q/A ability. In this paper, we propose SciEval, a comprehensive and multi-disciplinary evaluation benchmark to address these issues. Based on Bloom's taxonomy, SciEval covers four dimensions to systematically evaluate scientific research ability. In particular, we design a "dynamic" subset based on scientific principles to prevent evaluation from potential data leakage. Both objective and subjective questions are included in SciEval. These characteristics make SciEval a more effective benchmark for scientific research ability evaluation of LLMs. Comprehensive experiments on most advanced LLMs show that, although GPT-4 achieves SOTA performance compared to other LLMs, there is still substantial room for improvement, especially for dynamic questions. The codes and data are publicly available on https://github.com/OpenDFM/SciEval.

Downloads

Next from AAAI 2024

Efficient Online Crowdsourcing with Complex Annotations

Similar lecture

Generalization or Memorization: Data Contamination and Trustworthy Evaluation for Large Language Models

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from AAAI 2024

Efficient Online Crowdsourcing with Complex Annotations

Similar lecture

Generalization or Memorization: Data Contamination and Trustworthy Evaluation for Large Language Models

Downloads