Large Language Models (LLMs) demonstrate their reasoning ability through chain-of-thought (CoT) generation. However, LLM’s autoregressive decoding may limit the ability to revisit and refine earlier tokens in a holistic manner, which can also lead to inefficient exploration for d…
arXiv:2605.06040v1 Announce Type: cross Abstract: Although advances such as chain-of-thought, tree-of-thought or reinforcement learning have improved the performance of LLMs in reasoning and planning tasks, they are still brittle and have not achieved human-level performance in m…
arXiv cs.AI
TIER_1English(EN)·Jiahui Zhou, Dan Li, Boxin Li, Xiao Zhang, Erli Meng, Lin Li, Zhuomin Chen, Jian Lou, See-Kiong Ng·
arXiv:2602.07830v2 Announce Type: replace Abstract: Time series is a pervasive data type across various application domains, rendering the reasonable solving of diverse time series tasks a long-standing goal. Recent advances in large language models (LLMs), especially their reaso…
arXiv:2605.06241v1 Announce Type: new Abstract: Reinforcement learning has become the standard for improving reasoning in large language models, yet evidence increasingly suggests that RL does not teach new strategies; it redistributes probability mass over solutions the base mod…
Reinforcement learning has become the standard for improving reasoning in large language models, yet evidence increasingly suggests that RL does not teach new strategies; it redistributes probability mass over solutions the base model already contains. In this work, we ask: if RL…
Although advances such as chain-of-thought, tree-of-thought or reinforcement learning have improved the performance of LLMs in reasoning and planning tasks, they are still brittle and have not achieved human-level performance in many domains, and often suffer from high time and t…
arXiv:2605.04065v1 Announce Type: cross Abstract: Unsupervised reinforcement learning (RL) has emerged as a promising paradigm for enabling self-improvement in large language models (LLMs). However, existing unsupervised RL-based methods often lack the capacity to adapt to the mo…
arXiv:2605.04066v1 Announce Type: cross Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) is an essential paradigm that enhances the reasoning capabilities of Large Language Models (LLMs). However, existing methods typically rely on static policy optimization scheme…
arXiv:2605.02545v1 Announce Type: new Abstract: Large language models (LLMs) can generate syntactically valid optimization programs, yet often struggle to reliably choose an effective modeling strategy, leading to incorrect formulations and inefficient solver behavior. We propose…
arXiv:2605.02073v1 Announce Type: new Abstract: Mathematical reasoning is a key benchmark for large language models. Reinforcement learning is a standard post-training mechanism for improving the reasoning capabilities of large language models, yet performance remains sensitive t…
Large language models (LLMs) can generate syntactically valid optimization programs, yet often struggle to reliably choose an effective modeling strategy, leading to incorrect formulations and inefficient solver behavior. We propose SAGE, a strategy-aware framework that makes Mod…
arXiv cs.LG
TIER_1English(EN)·Zihan Lin, Xiaohan Wang, Jie Cao, Jiajun Chai, Li Wang, Xiaodong Lu, Wei Lin, Ran He, Guojun Yin·
arXiv:2605.00380v1 Announce Type: new Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) enhances reasoning of Large Language Models (LLMs) but usually exhibits limited generation diversity due to the over-incentivization of positive rewards. Although methods like Ne…
Mathematical reasoning is a key benchmark for large language models. Reinforcement learning is a standard post-training mechanism for improving the reasoning capabilities of large language models, yet performance remains sensitive to the design of the reward function that drives …
Mathematical reasoning is a key benchmark for large language models. Reinforcement learning is a standard post-training mechanism for improving the reasoning capabilities of large language models, yet performance remains sensitive to the design of the reward function that drives …
arXiv:2604.27713v1 Announce Type: new Abstract: The risks posed by AI features are increasing as they are rapidly integrated into software applications. In response, regulations and standards for safe and secure AI have been proposed. In this paper, we present an agentic framewor…
arXiv:2604.28031v1 Announce Type: new Abstract: When researchers iteratively refine ideas with large language models, do the models preserve fidelity to the original objective? We introduce DriftBench, a benchmark for evaluating constraint adherence in multi-turn LLM-assisted sci…
arXiv cs.AI
TIER_1English(EN)·Feiyu Wu, Xu Zheng, Zhuocheng Wang, Yi ming Dai, Hui Li·
arXiv:2604.28056v1 Announce Type: new Abstract: Large language models (LLMs) make reward design in reinforcement learning substantially more scalable, but generated rewards are not automatically reliable training objectives. Existing work has focused primarily on generating, evol…
arXiv cs.CL
TIER_1English(EN)·Byeongjin Kim, Gyuwan Kim, Seo Yeon Park·
arXiv:2601.11908v2 Announce Type: replace Abstract: Large language models (LLMs) struggle with reasoning over long contexts where relevant information is sparsely distributed. Although plan-and-execute frameworks mitigate this by decomposing tasks into planning and execution, the…
Reinforcement Learning with Verifiable Rewards (RLVR) enhances reasoning of Large Language Models (LLMs) but usually exhibits limited generation diversity due to the over-incentivization of positive rewards. Although methods like Negative Sample Reinforcement (NSR) mitigate this …
Large language models (LLMs) make reward design in reinforcement learning substantially more scalable, but generated rewards are not automatically reliable training objectives. Existing work has focused primarily on generating, evolving, or selecting reward candidates, while payi…
When researchers iteratively refine ideas with large language models, do the models preserve fidelity to the original objective? We introduce DriftBench, a benchmark for evaluating constraint adherence in multi-turn LLM-assisted scientific ideation. Across 2,146 scored benchmark …
When researchers iteratively refine ideas with large language models, do the models preserve fidelity to the original objective? We introduce DriftBench, a benchmark for evaluating constraint adherence in multi-turn LLM-assisted scientific ideation. Across 2,146 scored benchmark …
The risks posed by AI features are increasing as they are rapidly integrated into software applications. In response, regulations and standards for safe and secure AI have been proposed. In this paper, we present an agentic framework that constructs knowledge graphs (KGs) from AI…
arXiv cs.CL
TIER_1English(EN)·Zhenyu Zhao, Sander Land, Dan Bikel, Waseem Alshikh·
arXiv:2604.26355v1 Announce Type: new Abstract: Reasoning in Large Language Models incurs significant inference-time compute, yet the token-level information structure of reasoning traces remains underexplored. We observe that reasoning tokens split into two functional types: low…
Reasoning in Large Language Models incurs significant inference-time compute, yet the token-level information structure of reasoning traces remains underexplored. We observe that reasoning tokens split into two functional types: low-entropy \textit{structural} tokens (recurring p…
arXiv:2604.24827v1 Announce Type: new Abstract: Closed-source frontier labs do not disclose parameter counts, and the standard alternative -- inference economics -- carries $2\times$+ uncertainty from hardware, batching, and serving-stack assumptions external to the model. We exp…
arXiv:2604.25136v1 Announce Type: new Abstract: We propose Frictive Policy Optimization (FPO), a framework for learning language model policies that regulate not only what to say, but when and how to intervene in order to manage epistemic and normative risk. Unlike standard align…
arXiv:2510.19669v4 Announce Type: replace Abstract: Recent reasoning Large Language Models (LLMs) demonstrate remarkable problem-solving abilities but often generate long thinking traces whose utility is unclear. Our work aims to improve their efficiency, enabling them to reach h…
arXiv cs.CL
TIER_1English(EN)·Liaoyaqi Wang, Chunsheng Zuo, William Jurayj, Benjamin Van Durme, Anqi Liu·
arXiv:2604.23333v1 Announce Type: cross Abstract: Scaling test-time computation with reinforcement learning (RL) has emerged as a reliable path to improve large language models (LLM) reasoning ability. Yet, outcome-based reward often incentivizes models to be overconfident, leadi…
arXiv:2604.12373v5 Announce Type: replace Abstract: Humans use introspection to evaluate their understanding through private internal states inaccessible to external observers. We investigate whether large language models possess similar privileged knowledge about answer correctn…
arXiv:2604.23747v1 Announce Type: cross Abstract: Recent mixed-policy optimization methods for LLM reasoning that interleave or blend supervised and reinforcement learning signals report improvements over the standard SFT-then-RL pipeline. We show that numerous recently published…
arXiv cs.AI
TIER_1English(EN)·Hong Wang, Zhezheng Hao, Jian Luo, Chenxing Wei, Yao Shu, Lei Liu, Qiang Lin, Hande Dong, Jiawei Chen·
arXiv:2510.24832v2 Announce Type: replace Abstract: Using Reinforcement Learning with Verifiable Rewards (RLVR) to optimize Large Language Models (LLMs) can be conceptualized as progressively editing a query's "Reasoning Tree'. This process involves exploring nodes (tokens) and d…
arXiv:2509.21199v3 Announce Type: replace Abstract: Multi-Hop Question Answering (MHQA) requires integrating dispersed, interdependent evidence through sequential reasoning under noise. This task is challenging for LLMs as they have a finite per-pass output capacity, beyond which…
arXiv:2604.23270v1 Announce Type: new Abstract: Chain-of-Thought (CoT) prompting has emerged as a simple and effective way to elicit step-by-step solutions from large language models (LLMs). However, CoT reasoning can be unstable across runs on long, multi-step problems, leading …
arXiv cs.AI
TIER_1English(EN)·Junyan Cheng, Kyle Richardson, Peter Chin·
arXiv:2604.23072v1 Announce Type: new Abstract: Large language model (LLM) agents are increasingly tasked with complex real-world analysis (e.g., in financial forecasting, scientific discovery), yet their reasoning suffers from stochastic instability and lacks a verifiable, compo…
arXiv cs.CL
TIER_1English(EN)·Qibin Wang, Pu Zhao, Shaohan Huang, Fangkai Yang, Lu Wang, Furu Wei, Qingwei Lin, Saravan Rajmohan, Dongmei Zhang·
arXiv:2509.00084v2 Announce Type: replace-cross Abstract: Test-time scaling (TTS) has gained widespread attention for enhancing LLM reasoning. Existing approaches such as Best-of-N and majority voting are limited as their performance depends on the quality of candidate responses,…
arXiv:2508.01191v5 Announce Type: replace-cross Abstract: Chain-of-Thought (CoT) prompting has been shown to be effective in eliciting structured reasoning (i.e., CoT reasoning) from large language models (LLMs). Regardless of its popularity, recent studies expose its failures in…
We propose Frictive Policy Optimization (FPO), a framework for learning language model policies that regulate not only what to say, but when and how to intervene in order to manage epistemic and normative risk. Unlike standard alignment methods that optimize surface-level prefere…
arXiv cs.LG
TIER_1English(EN)·Md Muntaqim Meherab, Noor Islam S. Mohammad, Faiza Feroz·
arXiv:2603.10377v2 Announce Type: replace Abstract: Sparse autoencoders can localize where concepts live in language models, but not how they interact during multi-step reasoning. We propose Causal Concept Graphs (CCG): a directed acyclic graph over sparse, interpretable latent f…
Feature discovery from complex unstructured data is fundamentally a reasoning problem: it requires identifying abstractions that are predictive of a target outcome while avoiding leakage, proxies, and post-outcome signals. With the introduction of ever-improving Large Language Mo…
Reasoning in large language models is often treated as a monolithic capability, but its observed gains may arise from more basic operations. We study reasoning through two such primitives, recall and state-tracking, and ask whether hybrid architectures that combine attention-base…
Ahead of AI (Sebastian Raschka)
TIER_1English(EN)·Sebastian Raschka, PhD·
Large language models (LLMs) using chain-of-thought reasoning often waste substantial compute by producing long, incorrect responses. Abstention can mitigate this by withholding outputs unlikely to be correct. While most abstention methods decide to withhold outputs before or aft…
**HuggingFace** released **SmolLM3-3B**, a fully open-source small reasoning model with open pretraining code and data, marking a high point in open source models until **Olmo 3** arrives. **Grok 4** was launched with mixed reactions, while concerns about **Claude 4** nerfs and a…
