Formal Sciences

100 papers

Every entry carries a structured overview and the reproduction evidence collected so far. A paper is listed here when it carries the category as its primary field or as a secondary classification. Both are shown on the paper page.

Latest in Formal Sciences

Showing 41–60 of 100

Apr 22, 2026Formal Sciences260423.0041

Towards Certified Malware Detection: Provable Guarantees Against Evasion Attacks

Nandakrishna Giri, Asmitha K. A., Serena Nicolazzo, Antonino Nocera +1 more

This paper proposes a certifiably robust malware detection framework using randomized smoothing through feature ablation and noise injection to provide formal guarantees against adversarial evasion attacks. The approach achieves provable robustness without modifying the underlying ML classifier.

Apr 22, 2026Formal Sciences260423.0042

Decentralized Machine Learning with Centralized Performance Guarantees via Gibbs Algorithms

Yaiza Bermudez, Samir Perlaza, Iñaki Esnaola

This paper shows that decentralized machine learning can achieve the same performance as centralized learning by having clients share Gibbs measures (probability distributions over models) instead of raw data. The key innovation is using each client's Gibbs measure as a reference measure for the next client, effectively encoding prior information while preserving privacy.

Apr 22, 2026Formal Sciences260423.0043

Forecasting Individual NetFlows using a Predictive Masked Graph Autoencoder

Georgios Anyfantis, Pere Barlet-Ros

This paper proposes a Graph Neural Network model to predict network flow-level traffic (NetFlow) by representing network traffic as heterogeneous graphs with IP, Port, and Connection nodes. The approach demonstrates strong performance in identifying connection endpoints while maintaining competitive feature reconstruction compared to traditional forecasting baselines.

Apr 22, 2026Formal Sciences260423.0044

Temporal Difference Calibration in Sequential Tasks: Application to Vision-Language-Action Models

Shelly Francis-Meretzki, Mirco Mutti, Yaniv Romano, Aviv Tamar

This paper proposes temporal-difference calibration for vision-language-action models in robotics, connecting uncertainty calibration to reinforcement learning by showing that minimizing a sequential Brier score recovers the value function. The method improves calibration of task-success confidence in sequential decision-making tasks.

Apr 22, 2026Formal Sciences260423.0045

Surrogate Functionals for Machine-Learned Orbital-Free Density Functional Theory

Roman Remme, Fred A. Hamprecht

This paper introduces surrogate functionals, machine-learned energy functionals for orbital-free density functional theory that are optimized to produce correct ground-state densities rather than matching a physical reference. The approach requires only ground-state density data and achieves competitive accuracy while improving computational efficiency.

Apr 22, 2026Formal Sciences260423.0046

The Origin of Edge of Stability

Elon Litman

This paper explains why full-batch gradient descent on neural networks drives the largest Hessian eigenvalue toward the threshold 2/η (the Edge of Stability phenomenon) through a novel functional called edge coupling. The work provides a unified theoretical explanation for this self-regulating behavior that has previously resisted complete understanding.

Apr 22, 2026Formal Sciences260423.0047

DialToM: A Theory of Mind Benchmark for Forecasting State-Driven Dialogue Trajectories

Neemesh Yadav, Palakorn Achananuparp, Jing Jiang, Ee-Peng Lim

DialToM is a benchmark that tests whether LLMs truly understand Theory of Mind by evaluating both mental state prediction and the ability to forecast dialogue trajectories from those states. The study reveals that most LLMs can identify mental states but fail to use this understanding to predict realistic social interactions.

Apr 22, 2026Formal Sciences260423.0048

Unlocking the Forecasting Economy: A Suite of Datasets for the Full Lifecycle of Prediction Market: [Experiments \& Analysis]

Huaiyu Jia, Luofeng Zhou, Wentao Zhang, Lin William Cong +2 more

This paper introduces the first comprehensive dataset suite covering the complete lifecycle of decentralized prediction markets on Polymarket, integrating fragmented on-chain and off-chain data from October 2020 to March 2026. The unified dataset enables systematic analysis of market dynamics and supports downstream applications like outcome calibration and economic expectation reconstruction.

Apr 22, 2026Formal Sciences260423.0049

Scalable AI Inference: Performance Analysis and Optimization of AI Model Serving

Hung Cuong Pham, Fatih Gedikli

This paper evaluates and optimizes a BentoML-based AI inference system for scalable model serving, testing it under realistic workloads and applying multi-level optimizations to improve latency and throughput. The work bridges the gap between AI model research and practical deployment by providing concrete performance analysis and optimization strategies.

Apr 22, 2026Formal Sciences260423.0050

Calibrating conditional risk

Andrey Vasilyev, Yikai Wang, Xiaocheng Li, Guanting Chen

This paper introduces conditional risk calibration, the problem of estimating a model's expected loss conditional on input features, and shows it is equivalent to regression while being distinct from probability calibration. The work provides theoretical analysis and demonstrates practical applications in learning to defer frameworks.

Apr 20, 2026Formal Sciences260421.0038

MathNet: a Global Multimodal Benchmark for Mathematical Reasoning and Retrieval

Shaden Alshammari, Kevin Wen, Abrar Zainal, Mark Hamilton +4 more

MathNet is a large-scale, multilingual dataset of 30,676 Olympiad-level math problems from 47 countries spanning two decades, designed to benchmark both mathematical reasoning in generative models and mathematical retrieval in embedding systems. The benchmark reveals that even state-of-the-art models struggle with these problems, with top models achieving only 78.4% accuracy, and that retrieval quality significantly impacts retrieval-augmented generation performance.

Apr 20, 2026Formal Sciences260421.0039

Sessa: Selective State Space Attention

Liubomyr Horbatko

Sessa is a new sequence model that places attention inside a recurrent feedback path, enabling power-law memory decay instead of exponential or 1/length dilution. This architecture achieves superior long-context performance while remaining competitive on short sequences.

Apr 20, 2026Formal Sciences260421.0040

Bounded Ratio Reinforcement Learning

Yunke Ao, Le Chen, Bruce D. Lee, Assefa S. Wahd +4 more

This paper introduces Bounded Ratio Reinforcement Learning (BRRL), a theoretical framework that bridges the gap between trust region methods and PPO's clipped objective, leading to a new algorithm called Bounded Policy Optimization (BPO) that provides monotonic improvement guarantees while matching or exceeding PPO's performance. The framework also extends to Group-relative BPO (GBPO) for large language model fine-tuning.

Apr 20, 2026Formal Sciences260421.0041

When Can LLMs Learn to Reason with Weak Supervision?

Salman Rahman, Jingyan Shen, Anna Mordvina, Hamid Palangi +2 more

This paper investigates when reinforcement learning with verifiable rewards (RLVR) enables large language models to generalize under weak supervision (scarce data, noisy rewards, or self-supervised signals). The key finding is that models generalize when they exhibit prolonged pre-saturation training dynamics, which is predicted by reasoning faithfulness—the degree to which intermediate reasoning steps logically support final answers.

Apr 20, 2026Formal Sciences260421.0042

Back into Plato's Cave: Examining Cross-modal Representational Convergence at Scale

A. Sophia Koepke, Daniil Zverev, Shiry Ginosar, Alexei A. Efros

This paper challenges the Platonic Representation Hypothesis by showing that apparent alignment between vision and language models is an artifact of small-scale evaluation. When tested at scale with millions of samples and realistic many-to-many settings, cross-modal alignment degrades substantially, suggesting different modalities learn different representations of reality.

Apr 20, 2026Formal Sciences260421.0043

Revisiting Active Sequential Prediction-Powered Mean Estimation

Maria-Eleni Sfyraki, Jun-Kun Wang

This paper analyzes active sequential prediction-powered mean estimation, where labels are selectively queried and ML predictions fill in the gaps. The authors find that contrary to intuition, using a nearly constant query probability (ignoring uncertainty) often produces tighter confidence intervals than adaptive uncertainty-based querying.

Apr 20, 2026Formal Sciences260421.0044

Latent Phase-Shift Rollback: Inference-Time Error Correction via Residual Stream Monitoring and KV-Cache Steering

Manan Gupta, Dhruv Kumar

LPSR is an inference-time error correction method that monitors internal model activations to detect reasoning mistakes, then rolls back generation and steers the model using cached corrections—no training required. It improves an 8B model's math accuracy from 28.8% to 44.0%, outperforming prompted self-correction and even larger 70B models.

Apr 20, 2026Formal Sciences260421.0045

Benchmarking System Dynamics AI Assistants: Cloud Versus Local LLMs on CLD Extraction and Discussion

Terry Leitch

This paper benchmarks cloud and local large language models on two System Dynamics tasks: extracting causal loop diagrams and providing interactive coaching. The best local models match mid-tier cloud performance on diagram extraction (77%) but struggle with long-context error-fixing tasks, with backend implementation choices mattering more than quantization levels.

Apr 20, 2026Formal Sciences260421.0046

GSQ: Highly-Accurate Low-Precision Scalar Quantization for LLMs via Gumbel-Softmax Sampling

Alireza Dadgarnia, Soroush Tabesh, Mahdi Nikdan, Michael Helcig +2 more

GSQ is a new scalar quantization method for large language models that uses Gumbel-Softmax relaxation to jointly optimize grid assignments and scales, achieving accuracy comparable to complex vector quantization methods while remaining compatible with existing inference kernels. It successfully quantizes models to 2-3 bits per parameter and scales to trillion-parameter mixture-of-experts models.

Apr 20, 2026Formal Sciences260421.0047

A Note on TurboQuant and the Earlier DRIVE/EDEN Line of Work

Ran Ben-Basat, Yaniv Ben-Itzhak, Gal Mendelson, Michael Mitzenmacher +2 more

This note demonstrates that TurboQuant, a recent quantization method, is a suboptimal special case of the earlier EDEN/DRIVE quantization schemes. EDEN consistently outperforms TurboQuant across all tested scenarios, often by more than one bit of precision.