New model accounts for sequential dependence in LLM reliability.
problem Uncertainty in LLM reliability assessment due to sequential interactions.
method Extended Bayesian framework with Hidden Markov Model for sequential dependence.
result Ignoring sequential dependence leads to overconfident reliability estimates.
Study on limits of LLM-based multi-agent planning reliability.
problem Reliability limits of LLM-based multi-agent planning.
method Modeling LLM-based multi-agent architecture as a decision network, showing dominance by centralized Bayes decision maker.
result Optimizing multi-agent directed acyclic graphs under communication budget is equivalent to choosing a constrained experiment.
LLM evaluation suffers from systematic biases and lacks reliable positive judgments.
problem LLM evaluation suffers from systematic biases and lacks reliable positive judgments.
method Formulate LLM evaluation as a positive-unlabelled learning problem and propose a geometric auditing framework based on Partial Optimal Transport.
result Improved alignment with human preferences, increased robustness to presentation biases, and interpretable confidence estimates.
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.
R-AutoEval+ improves model evaluation efficiency and reliability using adaptive synthetic data.
problem Accurate model selection from AI candidates using real-world data is costly and impractical at scale.
method R-AutoEval+ uses adaptive prediction-powered inference to correct bias in autoevaluators while maintaining or improving sample efficiency.
result R-AutoEval+ provides finite-sample reliability guarantees and enhanced sample efficiency compared to conventional methods.
Study evaluates consistency of LLMs in binary text classification, providing systematic guidance.
problem Lack of reliable methods for evaluating large language model (LLM) binary text classification.
method Adapting psychometric principles, the study determines sample size requirements, develops metrics for invalid responses, and evaluates intra- and inter-rater reliability.
result LLMs demonstrated high intra-rater consistency, achieving perfect agreement on 90-98% of examples, with smaller models outperforming larger counterparts.
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.
The paper introduces multicalibration to improve confidence scores in LLMs.
problem Improving the reliability and interpretability of confidence scores for LLMs.
method Forming groupings of prompt/completion pairs correlated with correctness, using clustering and self-annotation. Developing multicalibration algorithms to reduce overfitting.
result Our techniques yield confidence scores that significantly improve calibration and accuracy compared to existing methods.
Unified framework for certifying LLM reliability without extra supervision.
problem Improving reliability of large language models without additional supervision.
method Unified framework using majority voting and Martingale Majority Certificate (MMC).
result Certifiable inference in LLMs with statistical guarantees and adaptive stopping rules.
Improved financial sentiment analysis using LLMs with retrieval augmentation.
problem Limited performance of traditional NLP models in financial sentiment analysis.
method Retrieval-augmented Large Language Models (LLMs) with instruction tuning.
result Achieved 15% to 48% performance gain in accuracy and F1 score.
AURA: Adaptive Uncertainty-aware Refinement for LLM-as-a-Judge Auditing
problem Auditing LLM-as-a-Judge decisions
method Adaptive uncertainty-aware refinement
result Human-consistent signal learning and evidence propagation
Framework assesses autograders' reliability and biases.
problem Mixed reliability and biases in autograders for LLM evaluation.
method Bayesian GLMs to model evaluation outcomes.
result Explicit quantification of scoring differences and biases.
Early stopping methods reduce unnecessary reasoning steps in LLMs by monitoring uncertainty signals.
problem LLMs sometimes generate unnecessary reasoning steps, especially under uncertainty.
method Statistically principled early stopping methods that monitor uncertainty signals during generation.
result Uncertainty-aware early stopping improves efficiency and reliability in LLM reasoning, especially in math reasoning.
CARE improves LLM aggregation by accounting for shared confounders.
problem LLM judges' correlated errors due to shared latent confounders.
method CARE explicitly models judges' scores as true quality and confounders.
result CARE reduces aggregation error by up to 26.8% across various benchmarks.
Optimal query allocation improves extractive QA efficiency with LLMs.
problem Efficiency and reliability in extractive question answering with LLMs.
method Learning-to-Defer framework that allocates queries to specialized models.
result Enhanced answer reliability with reduced computational overhead.
Benchmark assesses LLMs' causal inference skills, revealing significant limitations.
problem Lack of rigorous evaluation of LLMs' causal inference capabilities.
method CausalPitfalls benchmark with structured challenges and grading rubrics.
result Significant limitations in current LLMs' statistical causal inference.
Large language models struggle with causal relationships, leading to biases and hallucinations.
problem LLMs struggle with true causal relationships, leading to biases and hallucinations.
method Embed causality into LLMs training process at every stage.
result LLMs need to be trained to understand and apply causal knowledge, not just recite it.
New methods correct bias in LLM-as-a-Judge evaluations, but reliability depends on judge quality and model calibration.
problem Systematic bias in LLM-as-a-Judge evaluations using naive estimators.
method Analytical results, simulations, and real-data case study to diagnose reliability of corrected estimates.
result Corrected estimates, especially shared-calibration comparisons, can be unreliable under certain conditions.
Proposes integrating global and local entropy for more reliable LLMs.
problem Uncertainty in large language models (LLMs) leads to unreliable predictions.
method Measures global uncertainty from hidden-state matrices and local uncertainty from tokens, combining them via a multiplicative gate.
result Global-Local Uncertainty (GLU) outperforms unsupervised baselines across multiple models and benchmarks.
LLMs can help explain credit risk models but not autonomously.
problem Leveraging LLMs for post-hoc explainability in credit risk models.
method Comparison of LLM outputs with SHAP and coefficient-based attributions on three LMs.
result LLMs reliably preserve feature-importance rankings but poorly align with autonomous explanations.
Study evaluates five LLMs for financial report analysis, revealing performance differences and variability.
problem Lack of understanding in reliability, consistency, and transparency of LLMs in financial analysis.
method Human evaluation, automated similarity metrics, and behavioral diagnostics applied to five transformer-based LLMs over U.S. 10-K filings.
result No single LLM consistently dominates across all evaluation perspectives, highlighting variability and need for interpretability.
Study shows multilingual LLM calibration effects improve model confidence but not accuracy.
problem Improving multilingual language model calibration in low-resource settings.
method Analysis of two multilingual benchmarks using instruction-tuning and label smoothing.
result Model confidence increases in low-resource languages after instruction-tuning but accuracy improvements are marginal.
New method predicts nonfactuality in LLM responses using semantic isotropy.
problem Assessing trustworthiness of long-form LLM responses efficiently.
method Semantic isotropy of text embeddings on the unit sphere.
result Higher semantic isotropy correlates with lower factual consistency.
Research benchmarks LLMs in medical domain to reduce hallucinations.
problem Hallucinations in medical LLMs can lead to incorrect information.
method Developed Med-HALT dataset and testing methods.
result Significant performance differences among LLMs identified.
Corrects bias in LLM-as-a-judge evaluations using adaptive calibration.
problem Bias in LLM evaluations due to imperfect sensitivity and specificity.
method Plug-in framework with confidence intervals accounting for test and calibration dataset uncertainties.
result LML-based evaluation yields more reliable estimates than human-only evaluation.
New method uses model comparison signals to improve LLM evaluation accuracy.
problem Limited benchmark sizes and model stochasticity in evaluating LLMs' mathematical reasoning.
method Combines standard labeled outcomes with model comparison signals to design a statistically efficient evaluation framework.
result Semiparametric estimator achieves the semiparametric efficiency bound and substantially improves ranking accuracy.
New method certifies risks of LLM outputs, improving accuracy and reliability.
problem Uncertain and incorrect outputs from large language models.
method Information-lift certificates using PAC-Bayes bounds and skeleton design.
result Achieves 77.0% coverage at 2% risk, outperforming baselines.
LLMs struggle with financial reasoning but can outperform the market with human oversight.
problem Financial reasoning failures in LLM-generated stock market predictions.
method Evaluated four LLMs using three prompting strategies and compared to human oversight.
result LLMs require human oversight to fully realize their potential in financial markets.
We detect lookahead bias in LLM forecasts using a novel statistical method.
problem Detecting lookahead bias in LLM-generated economic forecasts.
method Developed a statistical procedure using date-only recall queries and estimated Lookahead Propensity (LAP).
result LLM forecasts are contaminated with lookahead bias, as indicated by a positive interaction between LAP and the forecast in accuracy regressions.
This work introduces RISE to explain LLMs more reliably by distinguishing essential context.
problem Identifying which context elements influence LLM outputs reliably.
method RISE (Redundancy-Insensitive Scoring of Explanation) method.
result RISE provides more robust explanations than traditional methods.
Paper presents a probabilistic model to improve LLM cascade performance.
problem Complexity of LLM cascades and their interaction error rates.
method Probabilistic model for joint performance distribution of LLMs.
result Improves area under the error-cost curve by 4.3% on average for cascades with k≥3 models.
PredictaBoard benchmarks LLM score predictors to assess their ability to anticipate errors.
problem Inconsistent performance of LLMs in common sense reasoning tasks.
method Collaborative benchmarking framework evaluating pairs of LLMs and assessors using rejection rate at different tolerance errors.
result Highlights the need to evaluate predictability alongside performance for safer AI systems.
G-Sim uses LLMs to build reliable simulators for complex systems.
problem Building robust simulators for critical domains like healthcare and logistics is challenging.
method Hybrid framework combining LLM-driven structural design and empirical calibration.
result G-Sim produces reliable, causally-informed simulators that handle non-differentiable and stochastic simulators.
This paper improves retrieval for LLMs in financial document Q&A.
problem Suboptimal text chunk retrieval by RAG causes inaccuracies in LLM responses.
method Sophisticated chunking techniques, query expansion, metadata annotations, re-ranking algorithms, and embedding fine-tuning.
result Enhanced retrieval quality improves LLM performance and reliability.
CausalGame benchmarks LLM agents' causal thinking in games.
problem Evaluating causal thinking in AI Scientists with LLMs.
method Interactive games with 14 scenarios incorporating selection bias, measurement error, and hidden confounders.
result None of the 30 LLM agents demonstrated reliable causal thinking, with the best model achieving only 68.0% survival.
Statsformer validates and adapts LLM-derived semantic priors for improved supervised learning.
problem Unreliable semantic priors from LLMs can degrade supervised learning performance.
method Adapts LLM-derived feature scores into a family of learner-specific prior-injection mechanisms, calibrating their influence using out-of-fold validation.
result Improves prediction performance by adaptively downweighting unreliable LLM priors, ensuring a guardrailed statistical learning system.
LLM Pro Finance Suite enhances financial NLP with instruction-tuned models.
problem Limited NLP capabilities for financial tasks in generalist models.
method Instruction-tuned large language models fine-tuned on financial data.
result Consistent improvement over state-of-the-art baselines in finance tasks.
New benchmarks focus on LLM risk in finance, not just accuracy.
problem Standard benchmarks ignore LLM financial risks, leading to unsafe deployment.
method Three-level agenda: model, workflow, and system stress-testing.
result Hidden weaknesses in LLMs are missed by traditional benchmarks.
Study examines flaws in probing LLMs' knowledge and introduces a new method.
problem Flaws in existing methods for probing the veracity of LLMs' internal knowledge.
method sAwMIL (Sparse-Aware Multiple-Instance Learning) combining multiple-instance learning with conformal prediction.
result LLMs encode a third type of signal distinct from true and false.
REALFIN benchmarks financial reasoning by removing implicit assumptions, revealing model weaknesses.
problem Models struggle when implicit assumptions are missing, leading to incorrect answers.
method Developed a bilingual benchmark that systematically removes essential premises from financial questions.
result General-purpose models over-commit, while finance-specialized models fail to identify missing premises.
MACI improves LLM factuality inference with higher retention and lower time cost.
problem Ensuring factuality in LLM responses for high-stakes domains.
method Reformulated conformal inference in a multiplicative filtering setting, leveraging ensembles for more accurate factuality scores and group-conditional calibration.
result MACI achieves higher retention and lower time cost compared to baselines, preserving validity through group-conditional calibration.
Evolutionary methods improve understanding of LLMs and their relationships.
problem Improving understanding of LLMs and their relationships.
method Relating weights to genotypes and output text to phenotypes using evolutionary methods.
result Estimated evolutionary trees reliably recover the topology of the ground-truth training tree.
TIM framework uses LLMs and domain experts to infer DeFi user transaction intents.
problem Challenges in understanding user intent in DeFi transactions due to complex interactions and opaque logs.
method TIM framework leverages a DeFi intent taxonomy, multi-agent LLM system, and a Meta-Level Planner.
result TIM significantly outperforms existing methods in inferring user transaction intents.
Auto-CEI improves LLM reasoning by balancing assertiveness and conservativeness.
problem Hallucinations and laziness in LLM reasoning tasks.
method Expert Iteration explores reasoning trajectories, guiding incorrect paths back on track and promoting appropriate 'I don't know' responses.
result Auto-CEI achieves superior alignment in logical reasoning, mathematics, and planning tasks.
Work shows hallucination detection by LLMs is impossible without expert feedback.
problem Detecting hallucinations in LLMs is theoretically impossible without expert-labeled feedback.
method Investigated hallucination detection using a theoretical framework inspired by language identification.
result Automated hallucination detection is impossible for most language collections without expert-labeled feedback.
RESTA defends LLMs against jailbreaking attacks by adding random noise to embeddings.
problem Vulnerability of LLMs to jailbreaking attacks that generate harmful outputs.
method Adds random noise to embedding vectors and aggregates during token generation.
result RESTA achieves superior robustness versus utility tradeoffs compared to baseline defenses.
SAEs struggle with feature consistency across runs, hindering MI reliability.
problem Inconsistency of learned SAE features across different training runs.
method Propose using the Pairwise Dictionary Mean Correlation Coefficient (PW-MCC) to measure feature consistency.
result High levels of feature consistency (0.80 for TopK SAEs on LLM activations) are achievable with appropriate architectural choices.
Bayesian approach quantifies uncertainty in LLM-based systems.
problem Uncertainty quantification in LLM-based systems, especially for high-stakes applications.
method Interpreting prompts as parameters in a Bayesian model, using MHLP for inference.
result Improvements in predictive accuracy and uncertainty quantification on various benchmarks.