Six state-of-the-art LLMs systematically prefer Standard American English over AAE continuations, and a training-free activation steering method reduces this bias 5-20x more than prompting while preserving fluency.
hub Mixed citations
Proceedings of the 2018
Mixed citation behavior. Most common role is background (67%).
49
Pith papers citing it
3,957
external citations · OpenAlex
Background
67% of classified citations
hub tools
citation-role summary
background 4
dataset 2
citation-polarity summary
representative citing papers
CheckMIABench converts LLMs with intermediate checkpoints into clean MIA testbeds by using pre- and post-checkpoint training data from the same distribution and evaluates published attacks on Pythia and OLMo models while releasing an open-source library.
Repetition rate mismatch between small-scale proxies and target budgets is the main reason data mixture experiments do not scale; a subsampling procedure that equalizes repetition rates recovers optimal mixtures from 1/16-scale experiments.
LLMs struggle to associate epistemic markers with stable internal confidence levels across distributions, even under model-centric interpretations, while maintaining somewhat consistent marker rankings.
A corpus-centric framework diagnoses scale, structure, overlap, metadata, and terminology properties across nine biomedical NER/EL corpora, showing substantial differences that common statistics fail to capture.
EdgeFlowerTune is a real-device benchmark that jointly assesses model quality and system costs for federated LLM fine-tuning on edge hardware using three protocols: Quality-under-Budget, Cost-to-Target, and Robustness.
ENMP prunes negative LoRA modules via evolutionary search to boost merging performance to new state-of-the-art levels across language and vision tasks.
Qualitative study of 19 practitioners reveals ten LLM product evaluation practices and introduces the results-actionability gap as a key barrier to turning findings into improvements.
Task prompt vectors, formed by subtracting random initialization from tuned soft prompts, support low-resource initialization and arithmetic combination across tasks on 12 NLU datasets while remaining independent of initialization seed on two model architectures.
RAG models set new state-of-the-art results on open-domain QA by retrieving Wikipedia passages and conditioning a generative model on them, while also producing more factual text than parametric baselines.
ALBERT reduces BERT parameters via embedding factorization and layer sharing, adds inter-sentence coherence pretraining, and reaches SOTA on GLUE, RACE, and SQuAD with fewer parameters than BERT-large.
On 2,520 programming tasks, matched Qwen general and coder models reliably raise Bloom cognitive demand but fail to lower it, so execution skill does not imply educational control.
Dialectal robustness and generation are dissociated in LLMs: benchmarks are driven by pretraining and SFT while alignment reshapes generation invisibly to benchmarks, and the method maximizing dialectal reward is least preferred by human evaluators.
Aggregating per-token pullback metrics via the Fréchet mean on the SPD manifold outperforms Euclidean mean pooling for sentence classification, with most of the gain attributable to geometric aggregation rather than learned encoder structure.
RLMF uses quality of model self-judgments to refine RL rankings and select training data, achieving SOTA faithful calibration while preserving accuracy and outperforming standard RL by up to 63%.
Pruning attention layers in five LLMs across eight datasets maintains accuracy but degrades faithfulness and calibration.
A post-hoc detection framework exploits generation-induced patterns in autoregressive image outputs to enable provenance tracing across multiple IAR models without altering the generation process.
SCOPE selects readable hidden layers, constructs conformal gates with IND calibration, and uses supermartingale e-processes to certify persistent service-boundary evidence, improving rejection over final-layer detectors across multiple LLMs and boundary conditions.
LLMs given only research questions from 1000 arXiv CS papers recommend a narrower set of methods than the original papers, with effective model-entity diversity dropping from 1232 to 59-96 and stronger agreement among LLMs than with papers.
MÖVE presents a new German-language benchmark evaluating 39 LLMs on performance and governance criteria using ten public-administration datasets.
Empirical benchmarks show distribution similarity between adaptation and pretraining data increases practical privacy leakage in DP-adapted LLMs at fixed theoretical guarantees, with LoRA providing strongest protection for OOD cases.
Explicitly disentangling semantic and positional streams in a Transformer encoder reveals that absolute positional representations collapse to a 2D document-structure manifold, attention heads specialize by role, and the approach improves linguistic probing performance on 49 of 65 phenomena.
SynLearner lets LLMs improve synthetic data generation on later tasks in a stream by learning reusable patterns and balancing quality with diversity from feedback on earlier tasks.
A new evaluation framework using MMD on Biber features shows LLMs deviate from human linguistic distributions across registers, with closest models varying by register rather than size.
citing papers explorer
-
LLMs Silently Correct African American English: Auditing and Mitigating Dialect Bias via Activation Steering
Six state-of-the-art LLMs systematically prefer Standard American English over AAE continuations, and a training-free activation steering method reduces this bias 5-20x more than prompting while preserving fluency.
-
CheckMIABench: Firm Foundations For Membership Inference Attacks on Language Models
CheckMIABench converts LLMs with intermediate checkpoints into clean MIA testbeds by using pre- and post-checkpoint training data from the same distribution and evaluates published attacks on Pythia and OLMo models while releasing an open-source library.
-
Repetition Mismatch: Why Data Mixture Experiments Don't Scale and How to Fix Them
Repetition rate mismatch between small-scale proxies and target budgets is the main reason data mixture experiments do not scale; a subsampling procedure that equalizes repetition rates recovers optimal mixtures from 1/16-scale experiments.
-
Can LLMs Use Linguistic Uncertainty Markers to Reliably Reflect Intrinsic Confidence?
LLMs struggle to associate epistemic markers with stable internal confidence levels across distributions, even under model-centric interpretations, while maintaining somewhat consistent marker rankings.
-
What Do Biomedical NER and Entity Linking Benchmarks Measure? A Corpus-Centric Diagnostic Framework
A corpus-centric framework diagnoses scale, structure, overlap, metadata, and terminology properties across nine biomedical NER/EL corpora, showing substantial differences that common statistics fail to capture.
-
EdgeFlowerTune: Evaluating Federated LLM Fine-Tuning Under Realistic Edge System Constraints
EdgeFlowerTune is a real-device benchmark that jointly assesses model quality and system costs for federated LLM fine-tuning on edge hardware using three protocols: Quality-under-Budget, Cost-to-Target, and Robustness.
-
Evolutionary Negative Module Pruning for Better LoRA Merging
ENMP prunes negative LoRA modules via evolutionary search to boost merging performance to new state-of-the-art levels across language and vision tasks.
-
Results-Actionability Gap: Understanding How Practitioners Evaluate LLM Products in the Wild
Qualitative study of 19 practitioners reveals ten LLM product evaluation practices and introduces the results-actionability gap as a key barrier to turning findings into improvements.
-
Task Prompt Vectors: Effective Initialization through Multi-Task Soft-Prompt Transfer
Task prompt vectors, formed by subtracting random initialization from tuned soft prompts, support low-resource initialization and arithmetic combination across tasks on 12 NLU datasets while remaining independent of initialization seed on two model architectures.
-
Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
RAG models set new state-of-the-art results on open-domain QA by retrieving Wikipedia passages and conditioning a generative model on them, while also producing more factual text than parametric baselines.
-
ALBERT: A Lite BERT for Self-supervised Learning of Language Representations
ALBERT reduces BERT parameters via embedding factorization and layer sharing, adds inter-sentence coherence pretraining, and reaches SOTA on GLUE, RACE, and SQuAD with fewer parameters than BERT-large.
-
From Execution to Education: A Bloom-Aligned Framework for Measuring Educational Control in LLMs
On 2,520 programming tasks, matched Qwen general and coder models reliably raise Bloom cognitive demand but fail to lower it, so execution skill does not imply educational control.
-
DiaLLM: An Investigation into the Robustness-Generation Gap in English Dialect Adaptation
Dialectal robustness and generation are dissociated in LLMs: benchmarks are driven by pretraining and SFT while alignment reshapes generation invisibly to benchmarks, and the method maximizing dialectal reward is least preferred by human evaluators.
-
Riemannian Geometry for Pre-trained Language Model Embeddings
Aggregating per-token pullback metrics via the Fréchet mean on the SPD manifold outperforms Euclidean mean pooling for sentence classification, with most of the gain attributable to geometric aggregation rather than learned encoder structure.
-
Reinforcement Learning with Metacognitive Feedback Elicits Faithful Uncertainty Expression in LLMs
RLMF uses quality of model self-judgments to refine RL rankings and select training data, achieving SOTA faithful calibration while preserving accuracy and outperforming standard RL by up to 63%.
-
Don't Go Breaking My LLM: The Impact of Pruning Attention Layers on Explanation Faithfulness and Confidence Calibration
Pruning attention layers in five LLMs across eight datasets maintains accuracy but degrades faithfulness and calibration.
-
Data Provenance for Image Auto-Regressive Generation
A post-hoc detection framework exploits generation-induced patterns in autoregressive image outputs to enable provenance tracing across multiple IAR models without altering the generation process.
-
SCOPE: Sequential Conformal Probing for Reliable OOD Rejection in LLM Services
SCOPE selects readable hidden layers, constructs conformal gates with IND calibration, and uses supermartingale e-processes to certify persistent service-boundary evidence, improving rejection over final-layer detectors across multiple LLMs and boundary conditions.
-
Thinking Like a Scientist? A Structural Study of LLM-Generated Research Methods
LLMs given only research questions from 1000 arXiv CS papers recommend a narrower set of methods than the original papers, with effective model-entity diversity dropping from 1232 to 59-96 and stronger agreement among LLMs than with papers.
-
M\"OVE: A Holistic LLM Benchmark for the German Public Sector
MÖVE presents a new German-language benchmark evaluating 39 LLMs on performance and governance criteria using ten public-administration datasets.
-
Benchmarking Empirical Privacy Protection for Adaptations of Large Language Models
Empirical benchmarks show distribution similarity between adaptation and pretraining data increases practical privacy leakage in DP-adapted LLMs at fixed theoretical guarantees, with LoRA providing strongest protection for OOD cases.
-
Give it Space! Explicit Disentangling of Positional and Semantic Representations in Encoders
Explicitly disentangling semantic and positional streams in a Transformer encoder reveals that absolute positional representations collapse to a 2D document-structure manifold, attention heads specialize by role, and the approach improves linguistic probing performance on 49 of 65 phenomena.
-
Make LLM Learn to Synthesize from Streaming Experiences through Feedback
SynLearner lets LLMs improve synthetic data generation on later tasks in a stream by learning reusable patterns and balancing quality with diversity from feedback on earlier tasks.
-
How Human-Like Are Large Language Models? A Register-Aware Linguistic Evaluation Framework
A new evaluation framework using MMD on Biber features shows LLMs deviate from human linguistic distributions across registers, with closest models varying by register rather than size.
-
GiLT: Augmenting Transformer Language Models with Dependency Graphs
GiLT augments Transformers with semantic dependency graphs by modulating attention to improve syntactic generalization while keeping perplexity competitive and enabling better finetuning on downstream tasks.
-
Measuring Five-Nines Reliability: Sample-Efficient LLM Evaluation in Saturated Benchmarks
Cross-entropy method sampling reduces inferences needed to estimate five-nines LLM reliability by up to 156x on parameterized GSM8K templates, revealing reliability differences hidden by saturated accuracy scores.
-
MC$^2$: Monte Carlo Correction for Fast Elliptic PDE Solving
MC² corrects low-budget Monte Carlo solutions for elliptic PDEs with a single-pass neural network to match the accuracy of 1000× more Monte Carlo samples while outperforming classical and learned baselines.
-
Extreme Weather Bench: A framework and benchmark for evaluation of high-impact weather
Extreme Weather Bench supplies standardized case studies, observational data, impact metrics, and code to evaluate weather models on high-impact hazards.
-
Compared to What? Baselines and Metrics for Counterfactual Prompting
Counterfactual prompting effects on LLMs are often indistinguishable from those caused by meaning-preserving paraphrases, causing most previously reported demographic sensitivities to disappear under proper statistical comparison.
-
Parameter-efficient Quantum Multi-task Learning
QMTL uses shared VQC encoding plus task-specific quantum ansatz heads to achieve linear parameter scaling with the number of tasks while matching or exceeding classical multi-task baselines on three benchmarks.
-
Preconditioned Norms: A Unified Framework for Steepest Descent, Quasi-Newton and Adaptive Methods
Preconditioned matrix norms unify steepest descent, quasi-Newton, and adaptive optimizers, revealing SGD, Adam, Muon, KL-Shampoo, SOAP, and SPlus as special cases and enabling new methods MuAdam and MuAdam-SANIA that are competitive in experiments.
-
HyperAdapt: Simple High-Rank Adaptation
HyperAdapt performs parameter-efficient fine-tuning by row- and column-wise diagonal scaling to induce high-rank updates with only n+m trainable parameters.
-
Should We Still Pretrain Encoders with Masked Language Modeling?
Controlled ablations of 38 models find MLM superior to CLM on representation benchmarks while CLM offers better data efficiency and stability; a biphasic CLM-then-MLM schedule is optimal under fixed compute and improves when initialized from pretrained CLM models.
-
How Good is Your Wikipedia? Auditing Data Quality for Low-resource and Multilingual NLP
The study filters non-English Wikipedia, reveals quality problems, proposes a 4-level ranking, and shows filtered data matches or beats raw data in language modeling with largest gains for lower-quality editions.
-
AGIEval: A Human-Centric Benchmark for Evaluating Foundation Models
AGIEval shows GPT-4 exceeding average human scores on SAT Math at 95% and Chinese college entrance English at 92.5%, while revealing weaker results on complex reasoning tasks.
-
Reproducibility Study of "AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models"
AlphaEdit's null-space editing guarantee holds only within its original models and ~3k-edit regime; it fails on Phi-3/Gemma-2 and degrades after ~5k edits while harming downstream and safety metrics.
-
More Yap Less Meaning: Uncovering Self-Improvement Behavior in SLMs
SLMs achieve only a 4.4% accuracy gain from self-generated hints on reasoning benchmarks, fail to semantically distinguish useful feedback, and perform worse with longer hints.
-
Position: State-of-the-Art Claims Require State-of-the-Art Evidence
SOTA claims based on aggregate benchmark scores frequently lack evidence for true model superiority beyond marginal mean improvements.
-
Survey-aware Machine Learning: A Guideline for Valid Population Health Inference based on Scoping Review
The authors introduce Survey-aware Machine Learning (SaML) as a nine-step guideline that integrates survey design metadata throughout the ML lifecycle to enable valid population inference from complex health surveys.
-
Fitting Is Not Enough: Smoothness in Extremely Quantized LLMs
Extremely quantized LLMs exhibit systematic smoothness degradation that reduces effective token candidates and degrades generation; a smoothness-preserving principle in PTQ and QAT delivers gains beyond numerical accuracy.
-
MIPIC: Matryoshka Representation Learning via Self-Distilled Intra-Relational and Progressive Information Chaining
MIPIC trains Matryoshka representations using self-distilled intra-relational alignment and progressive information chaining, yielding competitive results on STS, NLI, and classification tasks especially at low dimensions.
-
Toeplitz MLP Mixers are Low Complexity, Information-Rich Sequence Models
Toeplitz MLP Mixers replace attention with masked Toeplitz multiplications for sub-quadratic complexity while retaining more sequence information and outperforming on copying and in-context tasks.
-
Model-Agnostic Meta Learning for Class Imbalance Adaptation
HAMR combines meta-learning with hardness-aware weighting and neighborhood resampling to improve minority-class performance on imbalanced NLP datasets.
-
Decoding the Multimodal Maze: A Systematic Review on the Adoption of Explainability in Multimodal Attention-based Models
A systematic review of 55 papers finds explainability for multimodal attention-based models is dominated by attention-weight visualizations, while evaluation remains mostly qualitative and non-standardized.
-
Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference
ModernBERT is a new bidirectional encoder model achieving SOTA performance on diverse classification and retrieval benchmarks while offering superior speed and memory efficiency for long-context inference.
-
Benchmarked Yet Not Measured -- Generative AI Should be Evaluated Against Real-World Utility
Generative AI evaluation must shift from static benchmark scores to measuring sustained improvements in human capabilities within specific deployment contexts.
-
A Benchmark Suite of Reddit-Derived Datasets for Mental Health Detection
Four new Reddit-derived datasets for mental health detection tasks are presented with inter-annotator agreement above 0.8 and reported model F1 scores of 93-99%.
-
Investigating the Representation of Backchannels and Fillers in Fine-tuned Language Models
Fine-tuning on annotated English and Japanese dialogues improves clustering of backchannels and fillers and makes generated utterances closer to human ones.
-
Is Large Language Model Performance on Reasoning Tasks Impacted by Different Ways Questions Are Asked?
LLM accuracy on reasoning tasks differs significantly by question type, with step-by-step reasoning accuracy often uncorrelated to final answer selection.