- AgentDropoutV2: Optimizing Information Flow in Multi-Agent Systems via Test-Time Rectify-or-Reject Pruning
Yutong Wang, Siyuan Xiong, Xuebo Liu, Wenkang Zhou, Liang Ding · Feb 26, 2026 · Citations: 0
Automatic Metrics Multi Agent
While Multi-Agent Systems (MAS) excel in complex reasoning, they suffer from the cascading impact of erroneous information generated by individual participants.
- Why Diffusion Language Models Struggle with Truly Parallel (Non-Autoregressive) Decoding?
Pengxiang Li, Dilxat Muhtar, Lu Yin, Tianlong Chen, Shiwei Liu · Feb 26, 2026 · Citations: 0
Automatic Metrics
Across math reasoning benchmarks, NAP yields stronger performance under parallel decoding than DLMs trained on standard long CoT data, with gains growing as parallelism increases.
- InnerQ: Hardware-aware Tuning-free Quantization of KV Cache for Large Language Models
Sayed Mohammadreza Tayaranian Hosseini, Amir Ardakani, Warren J. Gross · Feb 26, 2026 · Citations: 0
Automatic Metrics
Our evaluation experiments on Llama models shows that InnerQ maintains a few-shot GSM8K performance comparable to non-quantized KV caches and surpasses prior KV cache quantization methods.
- A Decision-Theoretic Formalisation of Steganography With Applications to LLM Monitoring
Usman Anwar, Julianna Piskorz, David D. Baek, David Africa, Jim Weatherall · Feb 26, 2026 · Citations: 0
Automatic Metrics
Our central insight is that steganography creates an asymmetry in usable information between agents who can and cannot decode the hidden content (present within a steganographic signal), and this otherwise latent asymmetry can be inferred f
- NoRA: Breaking the Linear Ceiling of Low-Rank Adaptation via Manifold Expansion
Hung-Hsuan Chen · Feb 26, 2026 · Citations: 0
Automatic Metrics
On the SlimOrca benchmark, NoRA breaks this linear barrier: NoRA remarkably at rank 64 (PPL 3.89) outperforms LoRA at rank 512 (PPL 3.90), demonstrating superior spectral efficiency.
- Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching
Roy Miles, Aysim Toker, Andreea-Maria Oncescu, Songcen Xu, Jiankang Deng · Feb 26, 2026 · Citations: 0
Automatic Metrics Long Horizon
This modular pipeline separates exploration (diffusion) from evaluation and solution synthesis, avoiding monolithic unified hybrids while preserving broad search.
- Strategy Executability in Mathematical Reasoning: Leveraging Human-Model Differences for Effective Guidance
Weida Liang, Yiyou Sun, Shuyuan Nan, Chuang Li, Dawn Song · Feb 26, 2026 · Citations: 0
Automatic Metrics
Through a controlled analysis of paired human-written and model-generated solutions, we identify a systematic dissociation between usage and executability: human- and model-derived strategies differ in structured, domain-dependent ways, lea
- Dynamic Level Sets
Michael Stephen Fiske · Feb 26, 2026 · Citations: 0
Automatic Metrics
A mathematical concept is identified and analyzed that is implicit in the 2012 paper Turing Incomputable Computation, presented at the Alan Turing Centenary Conference (Turing 100, Manchester).
- Improving Parametric Knowledge Access in Reasoning Language Models
Melody Ma, John Hewitt · Feb 25, 2026 · Citations: 0
Automatic Metrics
We study reasoning for accessing world knowledge stored in a language model's parameters.
- Sparsity Induction for Accurate Post-Training Pruning of Large Language Models
Minhao Jiang, Zhikai Li, Xuewen Liu, Jing Zhang, Mengjuan Chen · Feb 25, 2026 · Citations: 0
Automatic Metrics
Large language models have demonstrated capabilities in text generation, while their increasing parameter scales present challenges in computational and memory efficiency.
- Duel-Evolve: Reward-Free Test-Time Scaling via LLM Self-Preferences
Sweta Karlekar, Carolina Zheng, Magnus Saebo, Nicolas Beltran-Velez, Shuyang Yu · Feb 25, 2026 · Citations: 0
Pairwise Preference Automatic Metrics
Building on this observation, we introduce Duel-Evolve, an evolutionary optimization algorithm that replaces external scalar rewards with pairwise preferences elicited from the same LLM used to generate candidates.
- Black-Box Reliability Certification for AI Agents via Self-Consistency Sampling and Conformal Calibration
Charafeddine Mouzouni · Feb 24, 2026 · Citations: 0
Automatic Metrics
We validate across five benchmarks, five models from three families, and both synthetic and real data.
- Aletheia tackles FirstProof autonomously
Tony Feng, Junehyuk Jung, Sang-hyun Kim, Carlo Pagano, Sergei Gukov · Feb 24, 2026 · Citations: 0
Automatic Metrics
We report the performance of Aletheia (Feng et al., 2026b), a mathematics research agent powered by Gemini 3 Deep Think, on the inaugural FirstProof challenge.
- Linear Reasoning vs. Proof by Cases: Obstacles for Large Language Models in FOL Problem Solving
Yuliang Ji, Fuchen Shen, Jian Wu, Qiujie Xie, Yue Zhang · Feb 24, 2026 · Citations: 0
Automatic Metrics
To comprehensively evaluate the mathematical reasoning capabilities of Large Language Models (LLMs), researchers have introduced abundant mathematical reasoning datasets.
- ID-LoRA: Efficient Low-Rank Adaptation Inspired by Matrix Interpolative Decomposition
Xindian Ma, Rundong Kong, Peng Zhang, Ruoxiang Huang, Yongyu Jiang · Feb 24, 2026 · Citations: 0
Automatic Metrics
We evaluate ID-LoRA on five diverse benchmarks: Mathematical Reasoning, Code Generation, MMLU, CommonsenseQA, and Safety Alignment.
- ToolMATH: A Math Tool Benchmark for Realistic Long-Horizon Multi-Tool Reasoning
Hyeonje Choi, Jeongsoo Lee, Hyojun Lee, Jay-Yoon Lee · Feb 24, 2026 · Citations: 0
Simulation Env Long Horizon
We introduce \ToolMATH, a math-grounded benchmark that evaluates tool-augmented language models in realistic multi-tool environments where the output depends on calling schema-specified tools and sustaining multi-step execution.
- GATES: Self-Distillation under Privileged Context with Consensus Gating
Alex Stein, Furong Huang, Tom Goldstein · Feb 24, 2026 · Citations: 0
Automatic Metrics Long Horizon
Held-out in-domain accuracy under asymmetric evaluation improves from 46.0\% to 62.0\%, and average (maj@8) accuracy on public document-free math benchmarks improves from 20.2\% to 35.4\%.
- KNIGHT: Knowledge Graph-Driven Multiple-Choice Question Generation with Adaptive Hardness Calibration
Mohammad Amanlou, Erfan Shafiee Moghaddam, Yasaman Amou Jafari, Mahdi Noori, Farhan Farsi · Feb 23, 2026 · Citations: 0
Automatic Metrics
Results show that KNIGHT enables token- and cost-efficient generation from a reusable graph representation, achieves high quality across these criteria, and yields model rankings aligned with MMLU-style benchmarks, while supporting topic-sp
- Position: General Alignment Has Hit a Ceiling; Edge Alignment Must Be Taken Seriously
Han Bao, Yue Huang, Xiaoda Wang, Zheyuan Zhang, Yujun Zhou · Feb 23, 2026 · Citations: 0
Automatic Metrics
We take the position that the dominant paradigm of General Alignment, which compresses diverse human values into a single scalar reward, reaches a structural ceiling in settings with conflicting values, plural stakeholders, and irreducible
- Pyramid MoA: A Probabilistic Framework for Cost-Optimized Anytime Inference
Arindam Khaled · Feb 23, 2026 · Citations: 0
Automatic Metrics
In this work, we propose "Pyramid MoA", a hierarchical Mixture-of-Agents architecture that uses a lightweight Router to dynamically escalate queries only when necessary.
- Do LLMs and VLMs Share Neurons for Inference? Evidence and Mechanisms of Cross-Modal Transfer
Chenhang Cui, An Zhang, Yuxin Chen, Gelei Deng, Jingnan Zheng · Feb 22, 2026 · Citations: 0
Automatic Metrics Long Horizon
Across diverse mathematics and perception benchmarks, SNRF consistently enhances LVLM inference performance while preserving perceptual capabilities.
- Think$^{2}$: Grounded Metacognitive Reasoning in Large Language Models
Abraham Paul Elenjical, Vivek Hruday Kavuri, Vasudeva Varma · Feb 21, 2026 · Citations: 0
Pairwise Preference Human Eval
We introduce a psychologically grounded metacognitive framework that operationalizes Ann Brown's regulatory cycle (Planning, Monitoring, and Evaluation) as a structured prompting architecture, and study its integration within a lightweight
- Watermarking LLM Agent Trajectories
Wenlong Meng, Chen Gong, Terry Yue Zhuo, Fan Zhang, Kecen Li · Feb 21, 2026 · Citations: 0
Automatic Metrics Long Horizon
LLM agents rely heavily on high-quality trajectory data to guide their problem-solving behaviors, yet producing such data requires substantial task design, high-capacity model generation, and manual filtering.
- VIRAASAT: Traversing Novel Paths for Indian Cultural Reasoning
Harshul Raj Surana, Arijit Maji, Aryan Vats, Akash Ghosh, Sriparna Saha · Feb 20, 2026 · Citations: 0
Automatic Metrics
Existing Cultural benchmarks are (i) Manually crafted, (ii) contain single-hop questions testing factual recall, and (iii) prohibitively costly to scale, leaving this deficiency largely unmeasured.
- SPQ: An Ensemble Technique for Large Language Model Compression
Jiamin Yao, Eren Gultepe · Feb 20, 2026 · Citations: 0
Automatic MetricsSimulation Env
Applied to LLaMA-2-7B, SPQ achieves up to 75% memory reduction while maintaining or improving perplexity (e.g., WikiText-2 5.47 to 4.91) and preserving accuracy on downstream benchmarks such as C4, TruthfulQA, and GSM8K.
- VeriSoftBench: Repository-Scale Formal Verification Benchmarks for Lean
Yutong Xin, Qiaochu Chen, Greg Durrett, Işil Dillig · Feb 20, 2026 · Citations: 0
Automatic Metrics
However, most benchmarks for LLM-based proof automation are drawn from mathematics in the Mathlib ecosystem, whereas proofs in software verification are developed inside definition-rich codebases with substantial project-specific libraries.
- Thinking by Subtraction: Confidence-Driven Contrastive Decoding for LLM Reasoning
Lexiang Tang, Weihao Gao, Bingchen Zhao, Lu Ma, Qiao jin · Feb 20, 2026 · Citations: 0
Automatic Metrics
Experiments show that CCD significantly improves accuracy across mathematical reasoning benchmarks while substantially reducing output length, with minimal KV-cache overhead.
- Gradient Regularization Prevents Reward Hacking in Reinforcement Learning from Human Feedback and Verifiable Rewards
Johannes Ackermann, Michael Noukhovitch, Takashi Ishida, Masashi Sugiyama · Feb 20, 2026 · Citations: 0
Automatic Metrics
Reinforcement Learning from Human Feedback (RLHF) or Verifiable Rewards (RLVR) are two key steps in the post-training of modern Language Models (LMs).
- TFL: Targeted Bit-Flip Attack on Large Language Model
Jingkai Guo, Chaitali Chakrabarti, Deliang Fan · Feb 19, 2026 · Citations: 0
Automatic Metrics
Large language models (LLMs) are increasingly deployed in safety and security critical applications, raising concerns about their robustness to model parameter fault injection attacks.
- ArXiv-to-Model: A Practical Study of Scientific LM Training
Anuj Gupta · Feb 19, 2026 · Citations: 0
Automatic Metrics
While frontier large language models demonstrate strong reasoning and mathematical capabilities, the practical process of training domain-specialized scientific language models from raw sources remains under-documented.
- BankMathBench: A Benchmark for Numerical Reasoning in Banking Scenarios
Yunseung Lee, Subin Kim, Youngjun Kwak, Jaegul Choo · Feb 19, 2026 · Citations: 0
Automatic Metrics Long Horizon
However, such errors have rarely been captured by existing benchmarks.
- RFEval: Benchmarking Reasoning Faithfulness under Counterfactual Reasoning Intervention in Large Reasoning Models
Yunseok Han, Yejoon Lee, Jaeyoung Do · Feb 19, 2026 · Citations: 0
Automatic Metrics
To operationalize this, we present RFEval, a benchmark of 7,186 instances across seven tasks that probes faithfulness via controlled, output-level counterfactual interventions.
- Training Large Reasoning Models Efficiently via Progressive Thought Encoding
Zeliang Zhang, Xiaodong Liu, Hao Cheng, Hao Sun, Chenliang Xu · Feb 18, 2026 · Citations: 0
Automatic Metrics
Experiments on three models, including Qwen2.5-3B-Instruct, Qwen2.5-7B-Instruct, and DeepSeek-R1-Distill-Llama-8B, on six widely used challenging mathematical benchmarks show consistent gains: our method achieves +19.3% improvement over LoR
- Utility-Preserving De-Identification for Math Tutoring: Investigating Numeric Ambiguity in the MathEd-PII Benchmark Dataset
Zhuqian Zhou, Kirk Vanacore, Bakhtawar Ahtisham, Jinsook Lee, Doug Pietrzak · Feb 18, 2026 · Citations: 0
Automatic Metrics
To address this challenge, we investigate the "numeric ambiguity" problem and introduce MathEd-PII, the first benchmark dataset for PII detection in math tutoring dialogues, created through a human-in-the-loop LLM workflow that audits upstr
- From Growing to Looping: A Unified View of Iterative Computation in LLMs
Ferdinand Kapl, Emmanouil Angelis, Kaitlin Maile, Johannes von Oswald, Stefan Bauer · Feb 18, 2026 · Citations: 0
Automatic Metrics
Looping, reusing a block of layers across depth, and depth growing, training shallow-to-deep models by duplicating middle layers, have both been linked to stronger reasoning, but their relationship remains unclear.
- Learning to Learn from Language Feedback with Social Meta-Learning
Jonathan Cook, Diego Antognini, Martin Klissarov, Claudiu Musat, Edward Grefenstette · Feb 18, 2026 · Citations: 0
Automatic Metrics
They are rarely proactive in soliciting this feedback, even when faced with ambiguity, which can make their dialogues feel static, one-sided, and lacking the adaptive qualities of human conversation.
- Recursive Concept Evolution for Compositional Reasoning in Large Language Models
Sarim Chaudhry · Feb 17, 2026 · Citations: 0
Automatic Metrics
Large language models achieve strong performance on many complex reasoning tasks, yet their accuracy degrades sharply on benchmarks that require compositional reasoning, including ARC-AGI-2, GPQA, MATH, BBH, and HLE.
- STAPO: Stabilizing Reinforcement Learning for LLMs by Silencing Rare Spurious Tokens
Shiqi Liu, Zeyu He, Guojian Zhan, Letian Tao, Zhilong Zheng · Feb 17, 2026 · Citations: 0
Automatic Metrics
Across six mathematical reasoning benchmarks using Qwen 1.7B, 8B, and 14B base models, STAPO consistently demonstrates superior entropy stability and achieves an average performance improvement of 7.13% ($ρ_{\mathrm{T}}$=1.0, top-p=1.0) and
- RUVA: Personalized Transparent On-Device Graph Reasoning
Gabriele Conte, Alessio Mattiace, Gianni Carmosino, Potito Aghilar, Giovanni Servedio · Feb 17, 2026 · Citations: 0
Automatic Metrics
We propose Ruva, the first "Glass Box" architecture designed for Human-in-the-Loop Memory Curation.
- Prescriptive Scaling Reveals the Evolution of Language Model Capabilities
Hanlin Zhang, Jikai Jin, Vasilis Syrgkanis, Sham Kakade · Feb 17, 2026 · Citations: 0
Automatic Metrics
Using large scale observational evaluations with 5k observational and 2k newly sampled data on model performance, we estimate capability boundaries, high conditional quantiles of benchmark scores as a function of log pre training FLOPs, via
- Weight space Detection of Backdoors in LoRA Adapters
David Puertolas Merenciano, Ekaterina Vasyagina, Raghav Dixit, Kevin Zhu, Ruizhe Li · Feb 16, 2026 · Citations: 0
Automatic Metrics
We evaluate the method on 500 LoRA adapters -- 400 clean, and 100 poisoned for Llama-3.2-3B on instruction and reasoning datasets: Alpaca, Dolly, GSM8K, ARC-Challenge, SQuADv2, NaturalQuestions, HumanEval, and GLUE dataset.
- Scaling Beyond Masked Diffusion Language Models
Subham Sekhar Sahoo, Jean-Marie Lemercier, Zhihan Yang, Justin Deschenaux, Jingyu Liu · Feb 16, 2026 · Citations: 0
Automatic Metrics
Among discrete diffusion approaches, Masked diffusion currently dominates, largely driven by strong perplexity on language modeling benchmarks.
- Cold-Start Personalization via Training-Free Priors from Structured World Models
Avinandan Bose, Shuyue Stella Li, Faeze Brahman, Pang Wei Koh, Simon Shaolei Du · Feb 16, 2026 · Citations: 0
Pairwise Preference Automatic Metrics
Cold-start personalization requires inferring user preferences through interaction when no user-specific historical data is available.
- Unlocking Reasoning Capability on Machine Translation in Large Language Models
Sara Rajaee, Sebastian Vincent, Alexandre Berard, Marzieh Fadaee, Kelly Marchisio · Feb 16, 2026 · Citations: 0
Critique Edit Automatic Metrics Long Horizon
We systematically evaluate several open- and closed-weights RLMs on the WMT24++ benchmark and find that enabling explicit reasoning consistently degrades translation quality across languages and models.
- Cognitive networks reconstruct mindsets about STEM subjects and educational contexts in almost 1000 high-schoolers, University students and LLM-based digital twins
Francesco Gariboldi, Emma Franchino, Edith Haim, Gianluca Lattanzi, Alessandro Grecucci · Feb 16, 2026 · Citations: 0
Automatic Metrics
Human networks show greater overlapping between mathematics and anxiety than GPT-oss.
- Learning beyond Teacher: Generalized On-Policy Distillation with Reward Extrapolation
Wenkai Yang, Weijie Liu, Ruobing Xie, Kai Yang, Saiyong Yang · Feb 12, 2026 · Citations: 0
Expert Verification Automatic Metrics
On-policy distillation (OPD), which aligns the student with the teacher's logit distribution on student-generated trajectories, has demonstrated strong empirical gains in improving student performance and often outperforms off-policy distil
- Search or Accelerate: Confidence-Switched Position Beam Search for Diffusion Language Models
Mingyu Cao, Alvaro H. C. Correia, Christos Louizos, Shiwei Liu, Lu Yin · Feb 11, 2026 · Citations: 0
Automatic Metrics
Across mathematical reasoning and code generation benchmarks (GSM8K, MBPP, HumanEval) on Dream-7B and LLaDA-8B, SOAR improves generation quality while maintaining competitive inference speed, offering a practical way to balance quality and
- Step 3.5 Flash: Open Frontier-Level Intelligence with 11B Active Parameters
Ailin Huang, Ang Li, Aobo Kong, Bin Wang, Binxing Jiao · Feb 11, 2026 · Citations: 0
Pairwise Preference Simulation Env Tool Use
We introduce Step 3.5 Flash, a sparse Mixture-of-Experts (MoE) model that bridges frontier-level agentic intelligence and computational efficiency.
- LLMs Know More About Numbers than They Can Say
Fengting Yuchi, Li Du, Jason Eisner · Feb 8, 2026 · Citations: 0
Automatic Metrics
Although state-of-the-art LLMs can solve math problems, we find that they make errors on numerical comparisons with mixed notation: "Which is larger, $5.7 \times 10^2$ or $580$?" This raises a fundamental question: Do LLMs even know how big
- Accelerating Scientific Research with Gemini: Case Studies and Common Techniques
David P. Woodruff, Vincent Cohen-Addad, Lalit Jain, Jieming Mao, Song Zuo · Feb 3, 2026 · Citations: 0
Automatic Metrics
Based on these experiences, we extract common techniques for effective human-AI collaboration in theoretical research, such as iterative refinement, problem decomposition, and cross-disciplinary knowledge transfer.
- Proof-RM: A Scalable and Generalizable Reward Model for Math Proof
Haotong Yang, Zitong Wang, Shijia Kang, Siqi Yang, Wenkai Yu · Feb 2, 2026 · Citations: 0
Automatic Metrics
In this work, we design a *scalable* data-construction pipeline that, with minimal human effort, leverages LLMs to generate a large quantity of high-quality ``**question-proof-check**'' triplet data.
- What If We Allocate Test-Time Compute Adaptively?
Ahsan Bilal, Ahmed Mohsin, Muhammad Umer, Ali Subhan, Hassan Rizwan · Feb 1, 2026 · Citations: 0
Automatic Metrics Long Horizon
For each problem, the agent runs multiple inference iterations.
- Unmasking Reasoning Processes: A Process-aware Benchmark for Evaluating Structural Mathematical Reasoning in LLMs
Xiang Zheng, Weiqi Zhai, Wei Wang, Boyu Yang, Wenbo Li · Jan 31, 2026 · Citations: 0
Automatic Metrics Multi Agent
Recent large language models (LLMs) achieve near-saturation accuracy on many established mathematical reasoning benchmarks, raising concerns about their ability to diagnose genuine reasoning competence.
- Group Representational Position Encoding
Yifan Zhang, Zixiang Chen, Yifeng Liu, Zhen Qin, Huizhuo Yuan · Dec 8, 2025 · Citations: 0
Automatic Metrics
We present GRAPE (Group Representational Position Encoding), a unified framework for positional encoding based on group actions.
- CDLM: Consistency Diffusion Language Models For Faster Sampling
Minseo Kim, Chenfeng Xu, Coleman Hooper, Harman Singh, Ben Athiwaratkun · Nov 24, 2025 · Citations: 0
Automatic Metrics
The full training and evaluation code is available at https://github.com/SqueezeAILab/CDLM.
- Long Grounded Thoughts: Synthesizing Visual Problems and Reasoning Chains at Scale
David Acuna, Chao-Han Huck Yang, Yuntian Deng, Jaehun Jung, Ximing Lu · Nov 7, 2025 · Citations: 0
Pairwise Preference Automatic Metrics
We introduce a framework able to synthesize vision-centric problems spanning diverse levels of complexity, and the resulting dataset with over 1M high-quality problems including: reasoning traces, preference data, and instruction prompts su
- A Proof of Learning Rate Transfer under $μ$P
Soufiane Hayou · Nov 3, 2025 · Citations: 0
Automatic Metrics
We provide the first proof of learning rate transfer with width in a linear multi-layer perceptron (MLP) parametrized with $μ$P, a neural network parameterization designed to ``maximize'' feature learning in the infinite-width limit.
- PARL: Prompt-based Agents for Reinforcement Learning
Yarik Menchaca Resendiz, Roman Klinger · Oct 24, 2025 · Citations: 0
Simulation Env
However, limited work evaluates LLMs as agents in reinforcement learning (RL) tasks (e.g., playing games), where learning occurs through interaction with an environment and a reward system.
- FinAuditing: A Financial Taxonomy-Structured Multi-Document Benchmark for Evaluating LLMs
Yan Wang, Keyi Wang, Shanshan Yang, Jaisal Patel, Jeff Zhao · Oct 10, 2025 · Citations: 0
Automatic Metrics
We introduce FinAuditing, a taxonomy-aligned, structure-aware benchmark built from real XBRL filings.
- Slm-mux: Orchestrating small language models for reasoning
Chenyu Wang, Zishen Wan, Hao Kang, Emma Chen, Zhiqiang Xie · Oct 6, 2025 · Citations: 0
Automatic Metrics
Additional experiments show that the core principle of SLM-MUX extends to open-ended generation tasks (e.g., HumanEval) and benefits other model classes, including frontier LLMs and domain-specific fine-tuned SLMs.
