Hierarchical framework for model evaluation on leaderboards
problem Uncertainty and variability in model performance across tasks
method Hierarchical framework with task-level and leaderboard-level rank prediction intervals
result Statistically valid and informative model rank intervals
The leaderboard in machine learning competitions is a tool to show the performance of various participants and to compare them. However, the leaderboard quickly becomes no longer accurate, due to hack or overfitting. This article gives two pieces of advice to prevent easy hack or overfitting. By following these advice,…
Representation learning, especially which by using deep learning, has been widely applied in classification. However, how to use limited size of labeled data to achieve good classification performance with deep neural network, and how can the learned features further improve classification remain indefinite. In this pa…
Large language models correlate in errors, even with different architectures and providers.
problem Lack of empirical evidence on whether different large language models differ meaningfully.
method Empirical evaluation of over 350 large language models using two leaderboards and a resume-screening task.
result Large language models have substantial correlation in errors, even with distinct architectures and providers.
Data competitions rely on real-time leaderboards to rank competitor entries and stimulate algorithm improvement. While such competitions have become quite popular and prevalent, particularly in supervised learning formats, their implementations by the host are highly variable. Without careful planning, a supervised lea…
Adversarial testing methods based on Projected Gradient Descent (PGD) are widely used for searching norm-bounded perturbations that cause the inputs of neural networks to be misclassified. This paper takes a deeper look at these methods and explains the effect of different hyperparameters (i.e., optimizer, step size an…
LLMs show biases in investment analysis, leading to unreliable recommendations.
problem LLMs face conflicts between pre-trained knowledge and real-time market data, leading to biases in investment analysis.
method Experimental framework to investigate emergent behaviors in LLMs, analyzing sector, size, and momentum biases.
result Distinct, model-specific biases observed, including a tendency to prefer technology stocks, large-cap stocks, and contrarian strategies.
Squirrel switches between optimizers for better performance.
problem Finding the best optimizer for a given problem.
method Switches between different optimizers based on performance.
result Improves performance on various problems.
Improved vehicle motion prediction with uncertainty estimation.
problem Robust motion prediction for autonomous vehicles, especially under distributional shift.
method Presented an approach significantly improving the benchmark and taking 2nd place on the leaderboard.
result Significantly improved motion prediction and uncertainty measurement.
This paper challenges the current metrics used for evaluating long-term forecasting models.
problem Current metrics focus on pointwise error reduction, ignoring structural properties.
method Proposes a multi-dimensional evaluation approach that includes statistical fidelity, structural coherence, and decision-level relevance.
result Current progress in forecasting may reflect specialization in benchmark configurations rather than deeper understanding of temporal dynamics.
New method improves Bayesian optimization by considering budget and prior quality.
problem Bayesian optimization lacks conditional treatment effects for specific contexts.
method Portable Regime Score (PRS) and RegimePlanner approach.
result Changing budget reverses optimization outcomes, and PRS predicts winners.
D2KLab's approach predicts tweet engagement using two stages.
problem Predicting user engagement with tweets.
method Two-stage approach: feature learning and ensemble XGBoost.
result Ranked 22 in the 2020 RecSys Challenge leaderboard.
ML Compass helps organizations choose AI models that balance utility, cost, and compliance.
problem Selecting AI models that meet user utility, deployment costs, and compliance requirements.
method Develops ML Compass, a framework for constrained optimization over a capability-cost frontier, using internal measures and empirical data.
result ML Compass produces deployment-aware recommendations that differ from capability-only rankings, clarifying trade-offs between capability, cost, and safety.
NIST's CTS Challenge evaluates speaker recognition using telephony data.
problem Improving speaker recognition accuracy using telephony data.
method Large-scale neural network architectures, extensive data augmentation, proprietary data, and fine-tuning.
result Remarkable improvements in speaker recognition performance.
We present SemEval-2019 Task 8 on Fact Checking in Community Question Answering Forums, which features two subtasks. Subtask A is about deciding whether a question asks for factual information vs. an opinion/advice vs. just socializing. Subtask B asks to predict whether an answer to a factual question is true, false or…
Many recent papers address reading comprehension, where examples consist of (question, passage, answer) tuples. Presumably, a model must combine information from both questions and passages to predict corresponding answers. However, despite intense interest in the topic, with hundreds of published papers vying for lead…
SCROLLS benchmarks long text NLP tasks, improving existing models.
problem Short NLP benchmarks ignore long texts; SCROLLS addresses this.
method Handpicked long-text datasets for summarization, QA, and inference tasks.
result Improvement potential on SCROLLS tasks, as indicated by initial baselines.
We are interested in learning causal relationships between pairs of random variables, purely from observational data. To effectively address this task, the state-of-the-art relies on strong assumptions regarding the mechanisms mapping causes to effects, such as invertibility or the existence of additive noise, which on…
ZeroSCROLLS benchmarks zero-shot natural language understanding over long texts.
problem Evaluate natural language understanding models over long texts without training data.
method Adapt six tasks from SCROLLS benchmark and add four new datasets, including novel aggregation tasks.
result Claude outperforms ChatGPT, and GPT-4 achieves highest average score.
Tiny benchmarks reduce LLM evaluation costs by using fewer examples.
problem Expensive evaluation of LLMs with tens of thousands of examples.
method Developed evaluation tools and tiny versions of popular benchmarks.
result Accurately estimate LLM performance with just 100 curated examples.
BIOMRC dataset improves MRC performance, especially for non-experts.
problem Improving machine reading comprehension in biomedical texts.
method Developed a new dataset and tested two neural models, along with a BERT-based model.
result BERT-based model outperforms all other methods and reaches expert-level accuracy.
CARROT optimizes LLM routing by choosing the cheapest and most accurate model.
problem Optimizing query routing to the most cost-effective LLM for a given task.
method CARROT uses cost and performance estimates to select the best LLM for each query.
result CARROT is minimax optimal, selecting the best LLM for any query.
Paper presents a model for identifying informative COVID-19 tweets.
problem Identifying informative COVID-19 tweets on Twitter.
method Leveraged transformers (RoBERTa, XLNet, BERTweet) trained in Semi-Supervised Learning (SSL) setting.
result Achieved F1 score of 0.9011 on test set, ranking 7th on leaderboard.
Camera and lidar are important sensor modalities for robotics in general and self-driving cars in particular. The sensors provide complementary information offering an opportunity for tight sensor-fusion. Surprisingly, lidar-only methods outperform fusion methods on the main benchmark datasets, suggesting a gap in the …
The Affective Behavior Analysis in-the-wild (ABAW) 2020 Competition is the first Competition aiming at automatic analysis of the three main behavior tasks of valence-arousal estimation, basic expression recognition and action unit detection. It is split into three Challenges, each one addressing a respective behavior t…
Bayesian optimization outperformed random search in machine learning hyperparameter tuning challenge.
problem Optimizing hyperparameters of machine learning models using derivative-free methods.
method Bayesian optimization vs. random search on real datasets.
result Bayesian optimization significantly outperformed random search in held-out objective functions.
A new framework evaluates LLMs by considering judge reliability.
problem Evaluating LLMs without ground truth labels can lead to biased results.
method Introduces judge-specific discrimination parameters and estimates model quality and judge reliability.
result Improves agreement with human preferences and produces calibrated uncertainty quantification.
Language model benchmarks often misrepresent true understanding, revealing vulnerabilities in evaluation methods.
problem Language model benchmarks fail to accurately reflect true language understanding and adaptability.
method Systematic analysis of NLP evaluation frameworks, identifying vulnerabilities in static benchmarks, human evaluation protocols, and LLM-as-judge frameworks.
result Current evaluation methods are unreliable and need improvement to accurately assess LLM performance.
The paper proposes using experts' insights in machine learning tasks.
problem Lack of reliance on experts' opinions in machine learning tasks.
method A three-step framework for integrating experts' insights into machine learning models.
result Proposed models achieve more stable and lower log loss results compared to top solutions.
A new framework evaluates large language models efficiently and accurately.
problem Evaluation of large language models is challenging due to stochasticity and heterogeneity of benchmarks.
method Interpretable and scalable framework based on Item Response Theory (IRT) and majorization-minimization principle.
result Our method achieves superior scalability and interpretability compared to existing approaches.
To access data stored in relational databases, users need to understand the database schema and write a query using a query language such as SQL. To simplify this task, text-to-SQL models attempt to translate a user's natural language question to corresponding SQL query. Recently, several generative text-to-SQL models …
KEMP predicts long-term trajectories for autonomous driving using keyframes.
problem Predicting future trajectories of road agents for autonomous driving.
method Keyframe-based hierarchical end-to-end deep learning framework.
result Ranked 1st on Waymo Open Motion Dataset Leaderboard.
PromptEval estimates LLM performance across many prompts, improving reproducibility.
problem Limited prompt templates affect LLM benchmark reproducibility.
method Estimates performance distribution across many prompts using borrowed strength.
result PromptEval accurately estimates performance quantiles with practical budget.
OGB provides diverse graph datasets for robust ML research.
problem Challenges in scalable and robust graph machine learning.
method Unified evaluation protocol, diverse datasets, and automated pipeline.
result Significant scalability and generalization challenges identified.
TIP-Search optimizes market prediction accuracy and timeliness under uncertain load.
problem Real-time market prediction requires accurate predictions before a deadline.
method Filters feasible models, dispatches workers, trades accuracy for deadline risk.
result Optimized pool achieves 0.991 timely accuracy and 0.994 raw accuracy.
Efficiently evaluate generative models at the prompt level using tensor factorization.
problem Fine-grained evaluations of generative models are costly and often misaligned with human judgment.
method Tensor factorization model that merges cheap autorater data with a small set of human gold-standard labels.
result The method provides accurate and tight confidence intervals for model performance.
Semantic code search is the task of retrieving relevant code given a natural language query. While related to other information retrieval tasks, it requires bridging the gap between the language used in code (often abbreviated and highly technical) and natural language more suitable to describe vague concepts and ideas…
The study uncovers latent capabilities of language models via causal representation learning.
problem Rigorous causal evaluations of language model capabilities are challenging due to confounding effects and computational costs.
method Proposes a causal representation learning framework to identify latent capability factors as causally interrelated after controlling for a common confounder (base model).
result Identifies a three-node linear causal structure explaining performance variations across 1500 models and six benchmarks.
SCR improves GNN training with consistency regularization.
problem Balancing labeled and unlabeled data in GNNs.
method Introduces two consistency regularization strategies: perturbed predictions and Mean Teacher.
result Enhances various GNNs to achieve better performance.
GIFT-Eval benchmarks time series forecasting models across diverse datasets.
problem Lack of comprehensive benchmarks for evaluating time series foundation models.
method Developed GIFT-Eval, a benchmark with 23 datasets, 177 million data points, and 144,000 time series.
result Promotes evaluation of foundation models across various domains and frequencies.
OpenDataVal benchmarks data valuation algorithms for diverse datasets.
problem Improving model performance and mitigating biases in training datasets.
method Unified benchmark framework for data valuation algorithms.
result No single algorithm performs uniformly best across all tasks.
Sloth predicts LLM performance using latent skills across families.
problem Variations in benchmark performance due to differences in training configurations and data processing across model families.
method Sloth uses publicly available benchmark data and assumes LLM performance is driven by latent skills influenced by model size and training tokens. It exploits correlations across benchmarks to provide accurate predictions.
result Sloth predicts LLM performance accurately and offers insights into scaling behaviors for complex tasks.
A new sparse benchmark metabench identifies key abilities from large benchmarks.
problem Redundancy and compression in existing benchmarks.
method Data from 5000+ LLMs to identify most informative items, distilling a sparse benchmark.
result Sparse benchmark metabench captures underlying abilities with high accuracy.
New framework selects high-quality pretraining data without training LLMs.
problem Slow progress in understanding pretraining data due to costly experiments.
method Statistical framework based on perplexity-benchmark correlations.
result Approach outperforms existing methods on multiple benchmarks.
BERT fine-tuning is unstable due to optimization issues, not forgetting or dataset size.
problem Stability of fine-tuning BERT-based models across different random seeds.
method Analysis of BERT, RoBERTa, and ALBERT fine-tuned on GLUE datasets, identifying optimization difficulties as the cause of instability.
result Fine-tuning instability is due to optimization difficulties leading to vanishing gradients, not forgetting or dataset size.
PowRL uses RL to manage power grids robustly, reducing overloads and maintaining power reliability.
problem Managing transient stability and preventing blackouts in power networks with uncertain generation and load demands.
method PowRL leverages a novel heuristic for overload management and RL-guided topology selection to ensure safe and reliable operation.
result PowRL outperforms other agents in L2RPN challenges, demonstrating robust performance in various scenarios.
TS-Fault benchmarks TSF models against structural faults.
problem Evaluating the robustness of time series forecasting models against structured events.
method TS-Fault uses parameterized fault scenarios with controllable difficulty.
result Three findings contradict common leaderboard intuition.
WILDS 2.0 expands benchmark datasets for unsupervised adaptation.
problem Leveraging unlabeled data for distribution shifts in real-world applications.
method Curated unlabeled data across various applications, tasks, and modalities.
result State-of-the-art methods perform poorly on WILDS datasets.