Copertina

Stable but Wrong: When More Data Degrades Scientific Conclusions

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Zhang, Zhipeng, Li, Kai
Natura:	Preprint
Pubblicazione:	2026
Soggetti:	Machine Learning Artificial Intelligence
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

When Learning Rates Go Wrong: Early Structural Signals in PPO Actor-Critic
di: Fernández-Hernández, Alberto, et al.
Pubblicazione: (2026)

Meta-Cognitive Reinforcement Learning with Self-Doubt and Recovery
di: Zhang, Zhipeng, et al.
Pubblicazione: (2026)

When LLMs Learn to Be Consistently Wrong: A Multi-Model Study of Linear Representations of Synthetic Deception
di: Zolfaghari, Vahideh
Pubblicazione: (2026)

When is an Embedding Model More Promising than Another?
di: Darrin, Maxime, et al.
Pubblicazione: (2024)

Training instability in deep learning follows low-dimensional dynamical principles
di: Zhang, Zhipeng, et al.
Pubblicazione: (2026)

Instance-level Randomization: Toward More Stable LLM Evaluations
di: Li, Yiyang, et al.
Pubblicazione: (2025)

All AI Models are Wrong, but Some are Optimal
di: Anand, Akhil S, et al.
Pubblicazione: (2025)

A Scientific Machine Learning Approach for Predicting and Forecasting Battery Degradation in Electric Vehicles
di: Murgai, Sharv, et al.
Pubblicazione: (2024)

When Less is More: 8-bit Quantization Improves Continual Learning in Large Language Models
di: Zhang, Michael S., et al.
Pubblicazione: (2025)

Shape of Thought: When Distribution Matters More than Correctness in Reasoning Tasks
di: Chandra, Abhranil, et al.
Pubblicazione: (2025)

Stable but Wrong: An Inference Limit in Galactic Archaeology
di: Zhang, Zhipeng
Pubblicazione: (2026)

Thinking Wrong in Silence: Backdoor Attacks on Continuous Latent Reasoning
di: Parekh, Swapnil
Pubblicazione: (2026)

When Stability Fails: Hidden Failure Modes Of LLMS in Data-Constrained Scientific Decision-Making
di: Riasat, Nazia
Pubblicazione: (2026)

More Than Routing: Joint GPS and Route Modeling for Refine Trajectory Representation Learning
di: Ma, Zhipeng, et al.
Pubblicazione: (2024)

LLM-Based Scientific Equation Discovery via Physics-Informed Token-Regularized Policy Optimization
di: Wang, Boxiao, et al.
Pubblicazione: (2026)

When a Robot is More Capable than a Human: Learning from Constrained Demonstrators
di: Li, Xinhu, et al.
Pubblicazione: (2025)

Easy Problems That LLMs Get Wrong
di: Williams, Sean, et al.
Pubblicazione: (2024)

When Fewer Layers Break More Chains: Layer Pruning Harms Test-Time Scaling in LLMs
di: Wang, Keyu, et al.
Pubblicazione: (2025)

When More is Less: Understanding Chain-of-Thought Length in LLMs
di: Wu, Yuyang, et al.
Pubblicazione: (2025)

The Significance of Latent Data Divergence in Predicting System Degradation
di: Fernandes, Miguel, et al.
Pubblicazione: (2024)

SciHorizon-DataEVA: An Agentic System for AI-Readiness Evaluation of Heterogeneous Scientific Data
di: Liu, Dianyu, et al.
Pubblicazione: (2026)

When Good Equations Get Bad Scores: Improving Symbolic Regression Through Better Parameter Optimization
di: Wang, Boxiao, et al.
Pubblicazione: (2026)

Reinforcement Learning-based Feature Generation Algorithm for Scientific Data
di: Xiao, Meng, et al.
Pubblicazione: (2025)

LLMComp: A Language Modeling Paradigm for Error-Bounded Scientific Data Compression (Technical Report)
di: Li, Guozhong, et al.
Pubblicazione: (2025)

SciTS: Scientific Time Series Understanding and Generation with LLMs
di: Wu, Wen, et al.
Pubblicazione: (2025)

Prompts Generalize with Low Data: Non-vacuous Generalization Bounds for Optimizing Prompts with More Informative Priors
di: Madras, David, et al.
Pubblicazione: (2025)

Why and When LLM-Based Assistants Can Go Wrong: Investigating the Effectiveness of Prompt-Based Interactions for Software Help-Seeking
di: Khurana, Anjali, et al.
Pubblicazione: (2024)

Weak-to-Strong Elicitation via Mismatched Wrong Drafts
di: Deng, Wei
Pubblicazione: (2026)

Perplexity Cannot Always Tell Right from Wrong
di: Veličković, Petar, et al.
Pubblicazione: (2026)

Post-hoc Interpretability Illumination for Scientific Interaction Discovery
di: Zhang, Ling, et al.
Pubblicazione: (2024)

More Expressive Feedforward Layers: Part I. Token-Adaptive Mixing of Activations
di: Wang, Mingze, et al.
Pubblicazione: (2026)

When Models Know More Than They Say: Probing Analogical Reasoning in LLMs
di: McGovern, Hope, et al.
Pubblicazione: (2026)

OmniArch: Building Foundation Model For Scientific Computing
di: Chen, Tianyu, et al.
Pubblicazione: (2024)

Rethinking Reward Model Evaluation: Are We Barking up the Wrong Tree?
di: Wen, Xueru, et al.
Pubblicazione: (2024)

Toward a Team of AI-made Scientists for Scientific Discovery from Gene Expression Data
di: Liu, Haoyang, et al.
Pubblicazione: (2024)

Generalizing Dynamics Modeling More Easily from Representation Perspective
di: Wang, Yiming, et al.
Pubblicazione: (2026)

Where Did It Go Wrong? Attributing Undesirable LLM Behaviors via Representation Gradient Tracing
di: Li, Zhe, et al.
Pubblicazione: (2025)

Deciphering Scientific Reasoning Steps from Outcome Data for Molecule Optimization
di: Liu, Zequn, et al.
Pubblicazione: (2026)

TabAttackBench: A Benchmark for Adversarial Attacks on Tabular Data
di: He, Zhipeng, et al.
Pubblicazione: (2025)

Continual Learning for Smart City: A Survey
di: Yang, Li, et al.
Pubblicazione: (2024)