William Yang Wang · page 2
Papers
- RuleArena: A Benchmark for Rule-Guided Reasoning with LLMs in Real-World Scenarios (2025 · 21 citations)
Benchmarks of LLM instruction-following mostly check style constraints such as format or length, or they test simple first-order logic. RuleArena instead tests whether models can reason under real regulations written in natural language. The authors gathered 95 rules from three domains: - ALGO: Synthesizing Algorithmic Programs with LLM-Generated Oracle Verifiers (2023 · 19 citations)
LLMs handle code generation well when a task spells out exactly what to implement. They do much worse on competitive-programming problems, where the model must first choose an efficient algorithm, and their output still needs human checking. ALGO tackles both issues. It builds on an observation: writing a slow… - Aristotle: Mastering Logical Reasoning with A Logic-Complete Decompose-Search-Resolve Framework (2025 · 16 citations)
Large language models still struggle with logical reasoning. General search-based methods such as ToT and GoT ignore the formal structure of logic problems: they split tasks along surface wording, depend on unreliable LLM evaluators to choose paths, and make logical slips when solving sub-steps. The authors present… - AntiLeakBench: Preventing Data Contamination by Automatically Constructing Benchmarks with Updated Real-World Knowledge (2025 · 12 citations)
Large language models are usually tested on public, static benchmarks, and those test items can leak into the training data of later models, which inflates their scores. - PECO: Examining Single Sentence Label Leakage in Natural Language Inference Datasets through Progressive Evaluation of Cluster Outliers (2023 · 9 citations)
NLI is meant to be a pairwise task: the relation between premise and hypothesis should depend on both sentences together. - Simple Models for Word Formation in English Slang (2018 · 8 citations)
The paper models three kinds of irregular word formation in English slang: blends (such as "fantabulous" from "fantastic" and "fabulous"), clippings (such as "prof" from "professor"), and reduplicatives (such as "boo-boo"). - “Love ya, jerkface”: Using Sparse Log-Linear Models to Build Positive and Impolite Relationships with Teens (2012 · 7 citations)
The paper asks how a tutoring dialogue system could read and signal rapport with teenagers, given that closeness shows up through both praise and friendly rudeness. - Detecting Training Data of Large Language Models via Expectation Maximization (2026 · 5 citations)
This paper tackles membership inference for large language models: deciding whether a given text was part of a model's pre-training data. - InductionBench: LLMs Fail in the Simplest Complexity Class (2025 · 5 citations)
Most reasoning benchmarks for LLMs test deduction: the rules are given (axioms, programming syntax) and must be applied. - WildSci: Advancing Scientific Reasoning from In-the-Wild Literature (2026)
Reinforcement learning with verifiable rewards (RLVR) has driven LLM reasoning gains mostly in math and code.
This page in Markdown → https://md.research.elamur.ai/free/en/authors/william-yang-wang/page/2/