News without the noise
66 results for “GSM8K” · page 3 of 4
Research · Reinforcement learning
ReFlow: Replay-Driven Decoupling for Efficient RL Post-Training
ReFlow, a decoupled and adaptive system, achieves up to 2.06× throughput improvement over existing methods while maintaining model quality.
DAPO-Math-17K: 0.2% rollout worker idle ratio (as reported)
The University of Hong Kong, University of Science and Technology of China, Alibaba Group
Research · Language models
Compatibility Matters: A Framework for Token-Level Teacher Selection in On-Policy Distillation
A theoretical framework and new routing scheme yield up to 10 percentage point gains in pass@k across math, code, and reasoning tasks.
AIME 2025 × KernelBench L2: 29.3% joint fast0 pass@16 (as reported)
Princeton University, Amazon Annapurna Labs
Research · Agents
Learning from Interaction Experience via Localized Context Distillation
The authors propose a method that identifies and revises only critical spans in a student's trajectory, outperforming both off-policy and on-policy context distillation across math, code, and agent tasks.
AIME24: 61.3% accuracy (avg@12) (as reported)
Tencent, The Chinese University of Hong Kong, Zhejiang University
Research · Language models
Learning to Revise Reasoning with Segment-wise On-Policy Distillation
Replacing student reasoning segments with teacher redrafts boosts accuracy by 5.22% over baselines.
AIME25: 59.7% accuracy (as reported)
The Hong Kong University of Science and Technology (Guangzhou), University of Science and Technology of China, Alibaba Group
Research · Language models
Towards Better Exploration in Sequential Test-Time Scaling
A model-mixing intervention helps language models avoid premature convergence and find answers beyond those accessible through parallel sampling.
multiple reasoning benchmarks: 2.2 pts accuracy improvement (recursive self-aggregation) (as reported)
University of Oxford, MBZUAI, BBC R&D
Research · Safety & security
SafeEvo: Deciphering the Safety Alignment Mechanism and Evolution in Language Models
SafeEvo identifies sparse refusal circuits in base models, traces their structural drift across safety checkpoints, and uses them to improve fine-tuning.
HarmBench: 6.47 pts average ASR reduction (as reported)
The University of Hong Kong, Chinese Academy of Sciences, Information Engineering University
Research · Language models
ALoDLM: Adaptively Looped Diffusion Language Models
ALoDLM recurrently refines difficult token representations while committing easier predictions early, improving benchmark scores without giving up parallel decoding.
GSM8K: 93.8% accuracy (as reported)
University of Illinois Chicago, Amazon AGI, Korea University
Research · Reinforcement learning
Residual Advantage: Student-Relative Teacher Guidance for RL with Verifiable Rewards
Chen and Pang show that centering the teacher's probability residual within each response allows step-level credit redistribution while preserving verifier outcome labels, improving reasoning performance across multiple benchmarks.
Macro: 37.12% Avg@8 (as reported)
Harbin Engineering University, Tencent, Harbin Institute of Technology
Research · Efficiency & systems
Few Bits, One Law: Toward W2A4KV2
CanonQ uses fixed transforms and reusable codebooks, then jointly trains models to tolerate quantized weights, activations and attention caches.
GSM8K: 47.8% final-answer exact match (as reported)
Meta AI
Research · Language models
Rethinking Cross-Tokenizer On-Policy Distillation: From Alignment Coverage to Supervision Reliability
Strict token alignment already captures most learning signal; adding span-level supervision reduces accuracy due to weakly aligned gradients.
ALFWorld: 30.34% success rate (as reported)
Alibaba Cloud Computing
Research · Language models
Efficient Reasoning via Constrained Optimization in Latent Space
Hou et al. show that efficient reasoning steps cluster in latent space; a quadratic programming correction reduces overthinking without causing underthinking.
GSM8K: 95.4% accuracy (as reported)
Tsinghua University
Research · Reinforcement learning
Lexicographic Multi-Objective On-Policy Distillation
Method routes student rollouts to highest-priority unmet specialist and projects corrections to avoid interfering with higher-priority experts.
AIME 2024, AIME 2025, HMMT February 2025: 88.6% retained gain, reasoning correctness (as reported)
Cohere, Mila, Université de Montréal
Research · Reinforcement learning
Outcome-Guided On-Policy Self-Distillation
The authors propose OG-OPSD, which selects forward or reverse KL divergence based on answer correctness and shortens distillation on incorrect trajectories using teacher entropy, outperforming vanilla on-policy self-distillation.
AIME 2024: 77.5% Avg@12 (as reported)
Southeast University, Huawei Noah’s Ark Lab, Zhongguancun Academy
Research · Generative media
Simplex Diffusion Models
Simplex Diffusion Models refine probability distributions over tokens, producing competitive language samples and stronger code-generation results with fewer denoising steps.
GSM8K: 32.1% problems solved (as reported)
Google DeepMind, EPFL, UCL Gatsby
Research · Language models
Learning to Stop without Learning to Stop: Self-Supervised Confidence Training Improves Reasoning Efficiency
Self-supervised confidence fine-tuning reduced token use across four model families without length penalties or inference-time stopping rules.
AIME2025: 45.6 accuracy (as reported)
University of Maryland, AI Foundations, Capital One
Research · Efficiency & systems
DScale: Scaling Block-Diffusion Speculative Decoding with Adaptive Verification
DScale reallocates limited verification capacity across requests while retaining useful draft tokens and fixed-shape GPU graphs.
GSM8K, MATH500, HumanEval, MBPP: 43.9% geometric-mean throughput gain vs DFlash (as reported)
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Victoria University of Wellington
Research · Efficiency & systems
A Tilted Bowl Is Not a Slippery Slope: Compressing Looped Models
Kolawole et al. show that rounding error in looped models shifts the fixed point rather than accumulating, enabling predictable failure and recovery with few high-precision steps.
GSM8K: 3% accuracy (as reported)
Carnegie Mellon University, ML Collective
Research · Language models
Large Language Continuous Diffusion Models
Sigma scales continuous language diffusion to 3B and 8B parameters, using 16-dimensional token embeddings to support controllable sampling and few-step generation.
GSM8K: 89.18% accuracy (as reported)
NVIDIA, Cornell University, Georgia Institute of Technology
Research · Safety & security
LLM Persona Unlearning
A new benchmark and contrastive erasure technique suppress target personas while preserving utility across six models and five personas.
PersonaUnlearnBench: 94.2% TF (as reported)
University of Macau, Mila – Québec AI Institute, McGill University
Research · Language models
Composable Decoding on the Probability Simplex: Theory and Implementation
The authors formalise decoding as a regularised optimisation problem, recovering standard methods as special cases and enabling new decoders through compositional regularisers.
GPQA Diamond: 26.3% pass@1 (as reported)
Huawei Noah's Ark Lab, UCL Centre for AI