arXiv subjects

Youran Sun

Publications and source records attributed to Youran Sun.

29 records · Page 2Linked to original sources

arXiv2026

DRIFT: Detecting Representational Inconsistencies for Factual Truthfulness

LLMs often produce fluent but incorrect answers, yet detecting such hallucinations typically requires multiple sampling passes or post-hoc verification, adding significant latency and cost. We hypothesize that intermediate layers encode confidence signals that are lost in the final output layer, and propose a lightweight probe to read these signals directly from hidden states. The probe adds less than 0.1\% computational overhead and can run fully in parallel with generation, enabling hallucination detection before the answer is produced. Building on this, we develop an LLM router that answers confident queries immediately while delegating uncertain ones to stronger models. Despite its simplicity, our method achieves SOTA AUROC on 10 out of 12 settings across four QA benchmarks and three LLM families, with gains of up to 13 points over prior methods, and generalizes across dataset shifts without retraining.

cs.CL↗

arXiv2026

OptimAI: Optimization from Natural Language Using LLM-Powered AI Agents

Optimization plays a vital role in scientific research and practical applications. However, formulating a concrete optimization problem described in natural language into a mathematical form and selecting a suitable solver to solve the problem requires substantial domain expertise. We introduce OptimAI, a framework for solving Optimization problems described in natural language by leveraging LLM-powered AI agents, and achieve superior performance over current state-of-the-art methods. Our framework is built upon the following key roles: (1) a formulator that translates natural language problem descriptions into precise mathematical formulations; (2) a planner that constructs a high-level solution strategy prior to execution; and (3) a coder and a code critic capable of interacting with the environment and reflecting on outcomes to refine future actions. Ablation studies confirm that all roles are essential; removing the planner or code critic results in $5.8\times$ and $3.1\times$ drops in productivity, respectively. Furthermore, we introduce UCB-based debug scheduling to dynamically switch between alternative plans, yielding an additional $3.3\times$ productivity gain. Our design emphasizes multi-agent collaboration, and our experiments confirm that combining diverse models leads to performance gains. Our approach attains 88.1% accuracy on the NLP4LP dataset and 82.3% on the Optibench dataset, reducing error rates by 58% and 52%, respectively, over prior best results.

cs.CL↗

arXiv2025

FrontierCS: Evolving Challenges for Evolving Intelligence

We introduce FrontierCS, a benchmark of 156 open-ended problems across diverse areas of computer science, designed and reviewed by experts, including CS PhDs and top-tier competitive programming participants and problem setters. Unlike existing benchmarks that focus on tasks with known optimal solutions, FrontierCS targets problems where the optimal solution is unknown, but the quality of a solution can be objectively evaluated. Models solve these tasks by implementing executable programs rather than outputting a direct answer. FrontierCS includes algorithmic problems, which are often NP-hard variants of competitive programming problems with objective partial scoring, and research problems with the same property. For each problem we provide an expert reference solution and an automatic evaluator. Combining open-ended design, measurable progress, and expert curation, FrontierCS provides a benchmark at the frontier of computer-science difficulty. Empirically, we find that frontier reasoning models still lag far behind human experts on both the algorithmic and research tracks, that increasing reasoning budgets alone does not close this gap, and that models often over-optimize for generating merely workable code instead of discovering high-quality algorithms and system designs.

cs.LG↗

arXiv2025

Jailbreaking LLMs' Safeguard with Universal Magic Words for Text Embedding Models

The security issue of large language models (LLMs) has gained wide attention recently, with various defense mechanisms developed to prevent harmful output, among which safeguards based on text embedding models serve as a fundamental defense. Through testing, we discover that the output distribution of text embedding models is severely biased with a large mean. Inspired by this observation, we propose novel, efficient methods to search for **universal magic words** that attack text embedding models. Universal magic words as suffixes can shift the embedding of any text towards the bias direction, thus manipulating the similarity of any text pair and misleading safeguards. Attackers can jailbreak the safeguards by appending magic words to user prompts and requiring LLMs to end answers with magic words. Experiments show that magic word attacks significantly degrade safeguard performance on JailbreakBench, cause real-world chatbots to produce harmful outputs in full-pipeline attacks, and generalize across input/output texts, models, and languages. To eradicate this security risk, we also propose defense methods against such attacks, which can correct the bias of text embeddings and improve downstream performance in a train-free manner.

cs.CL↗

arXiv2025

LLMs Meet Finance: Fine-Tuning Foundation Models for the Open FinLLM Leaderboard

This paper investigates the application of large language models (LLMs) to financial tasks. We fine-tuned foundation models using the Open FinLLM Leaderboard as a benchmark. Building on Qwen2.5 and Deepseek-R1, we employed techniques including supervised fine-tuning (SFT), direct preference optimization (DPO), and reinforcement learning (RL) to enhance their financial capabilities. The fine-tuned models demonstrated substantial performance gains across a wide range of financial tasks. Moreover, we measured the data scaling law in the financial domain. Our work demonstrates the potential of large language models (LLMs) in financial applications.

cs.CL↗

arXiv2025

Phase Transitions in Large Language Models and the $O(N)$ Model

Large language models (LLMs) exhibit unprecedentedly rich scaling behaviors. In physics, scaling behavior is closely related to phase transitions, critical phenomena, and field theory. To investigate the phase transition phenomena in LLMs, we reformulated the Transformer architecture as an $O(N)$ model. Our study reveals two distinct phase transitions corresponding to the temperature used in text generation and the model's parameter size, respectively. The first phase transition enables us to estimate the internal dimension of the model, while the second phase transition is of \textit{higher-depth} and signals the emergence of new capabilities. As an application, the energy of the $O(N)$ model can be used to evaluate whether an LLM's parameters are sufficient to learn the training data.

cs.LG↗

arXiv2025

rStar-Math: Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinking

We present rStar-Math to demonstrate that small language models (SLMs) can rival or even surpass the math reasoning capability of OpenAI o1, without distillation from superior models. rStar-Math achieves this by exercising "deep thinking" through Monte Carlo Tree Search (MCTS), where a math policy SLM performs test-time search guided by an SLM-based process reward model. rStar-Math introduces three innovations to tackle the challenges in training the two SLMs: (1) a novel code-augmented CoT data sythesis method, which performs extensive MCTS rollouts to generate step-by-step verified reasoning trajectories used to train the policy SLM; (2) a novel process reward model training method that avoids naïve step-level score annotation, yielding a more effective process preference model (PPM); (3) a self-evolution recipe in which the policy SLM and PPM are built from scratch and iteratively evolved to improve reasoning capabilities. Through 4 rounds of self-evolution with millions of synthesized solutions for 747k math problems, rStar-Math boosts SLMs' math reasoning to state-of-the-art levels. On the MATH benchmark, it improves Qwen2.5-Math-7B from 58.8% to 90.0% and Phi3-mini-3.8B from 41.4% to 86.4%, surpassing o1-preview by +4.5% and +0.9%. On the USA Math Olympiad (AIME), rStar-Math solves an average of 53.3% (8/15) of problems, ranking among the top 20% the brightest high school math students. Code and data will be available at https://github.com/microsoft/rStar.

cs.CL↗

arXiv2024

Understand the Effectiveness of Shortcuts through the Lens of DCA

Difference-of-Convex Algorithm (DCA) is a well-known nonconvex optimization algorithm for minimizing a nonconvex function that can be expressed as the difference of two convex ones. Many famous existing optimization algorithms, such as SGD and proximal point methods, can be viewed as special DCAs with specific DC decompositions, making it a powerful framework for optimization. On the other hand, shortcuts are a key architectural feature in modern deep neural networks, facilitating both training and optimization. We showed that the shortcut neural network gradient can be obtained by applying DCA to vanilla neural networks, networks without shortcut connections. Therefore, from the perspective of DCA, we can better understand the effectiveness of networks with shortcuts. Moreover, we proposed a new architecture called NegNet that does not fit the previous interpretation but performs on par with ResNet and can be included in the DCA framework.

cs.LG↗

arXiv2024

Modularity of Vafa-Witten Partition Functions from SymTFT

The 6d (2,0) theory of $N$ M5 branes compactified on the product geometry $T^2\times S$, where $S$ is a Kähler 4-manifold, can be studied in two different limits. In one limit, the size of $T^2$ is taken to zero and together with a topological twist one arrives at the Vafa-Witten partition function on $S$. On the other hand, taking the size of $S$ to zero leads to a 2d $\mathcal{N}=(0,4)$ theory. This gives rise to a 2d-4d correspondence where the Vafa-Witten partition functions are identified with the characters of the 2d theory. In this paper, we test this conjecture for Hirzebruch and Del Pezzo surfaces by employing the technique of SymTFT to show that the modular transformation properties of the two sides match. Moreover, we construct modular invariant 2d absolute partition functions and verify that they are invariant under gauging of a discrete symmetry at the self-dual point in coupling space. This provides further hints for the presence of duality defects in the 2d SCFT.

hep-th↗

arXiv2024

Topological Defect Lines in bosonized Parafermionic CFTs

Topological defect lines (TDLs) are extended line operators which act on the Hilbert space of two-dimensional CFTs and satisfy non-trivial fusion algebras when forming junctions. Among the most interesting fusion algebras are the so-called Tambara-Yamagami (TY) fusion categories which are realized in (bosonized) Parafermionic CFTs. The corresponding TY[$\mathbb{Z}_k$]-categories have been explicitly realized for the cases $k=2$, $k=3$, and $k=4$ together with the action of the defect lines on the Hilbert space of the corresponding CFTs. For each of the cases, different methods have been used in the previous literature. In the current paper, we present a unified framework for finding the TDLs in bosonized Parafermionic CFTs. Our approach relies on combining several previously used methods and the definition of an extended $S$ matrix. We apply the method to the cases $k=2$ to $k=5$ to extract corresponding TDL fusion algebras.

hep-th↗

arXiv2021

Restricted Boltzmann machine representation for the groundstate and excited states of Kitaev Honeycomb model

In this work, the capability of restricted Boltzmann machines (RBMs) to find solutions for the Kitaev honeycomb model with periodic boundary conditions is investigated. The measured groundstate (GS) energy of the system is compared and, for small lattice sizes (e.g. $3 \times 3$ with $18$ spinors), shown to agree with the analytically derived value of the energy up to a deviation of $0.09\%$. Moreover, the wave-functions we find have $99.89\%$ overlap with the exact ground state wave-functions. Furthermore, the possibility of realizing anyons in the RBM is discussed and an algorithm is given to build these anyonic excitations and braid them for possible future applications in quantum computation. Using the correspondence between topological field theories in (2+1)d and 2d CFTs, we propose an identification between our RBM states with the Moore-Read state and conformal blocks of the $2$d Ising model.

cond-mat.dis-nn↗