Qwen3-0.6B mixes 4-bit decode with selective 8-bit steps
An entropy gate at a 16% escalation rate scores 62.7% on 150 GSM8K problems, using 0.87 of all-high bytes.
Stories per day · 30 days
2 RSS
Relateddiffusion-language-models2code-generation1dream1inference1language-modeling1language-models1llada1masked-diffusion1mixed-precision1openwebtext1quantization1qwen1schedule-distillation1self-conditioning1self-distillation1
16 stories, newest first · one card per story
Stories tagged #gsm8k
HiLo-RD improves low-budget masked diffusion language models
Self-distillation from large-window states raises GSM8K accuracy by 9.4 points and cuts mean readable length by about 25%.
dSpark distills unmasking schedules to speed diffusion language models
On GSM8K the paper reports 10.12× and 5.12× speedups on LLaDA-8B-Instruct and Dream-7B-Instruct while keeping task quality.
LAW bases language modeling entirely on self-conditioning
After TinyGSM training, LAW solves 42.2% of GSM8K at 1024 NFE versus 34.5% for uniform loopholing discrete diffusion.
Adversarial Cues in Decision Models Used as Judges: The Role of Request Presentation
A single colon raised LLM judge Jev's false acceptance of wrong answers from about 1% to 26%.
Denoising Hierarchical Representations: Joint Continuous Diffusion for Language Modeling
Hierarchical continuous diffusion models jointly denoise tokens and clusters, improving generative perplexity by over 20 points.
DP-ES: Differentially Private Evolution Strategies for Prompt Optimization
DP-ES privately optimizes full prompts and beats DP-OPT by 38.6 points on GSM8K under a tight privacy budget.
Representation-Space MMD for Diffusion Language Models
Representation-space MMD post-training improves diffusion language models without full sampling trajectories.
Sharpen Without Search: On-Policy Distillation of Sequence-Level Power Distribution
On-policy power distillation trains one-shot sharpened reasoning, lifting MATH500 by up to 23 points.
Reasoning with Continuous Latent Diffusion
Latent Flow Reasoning Models generate math and code solutions by iterative denoising in continuous latent space.
Why Deterministic PRM Guidance Underperforms in Discrete Diffusion Reasoning
Deterministic PRM guidance for discrete diffusion LLMs loses to simpler ORM reranking at matched compute.
Kyutai Releases Voice of Reason: A Speech-Native Model that Solves Spoken Math with Reinforcement Learning
Kyutai released two open-weight 9B speech models that raise spoken GSM8K accuracy from 27.3% to 77.1% using reinforcement learning.
3
Flash-dLLM: IO-Aware KV Caching and Parallel Decoding for Fast, Memory-Efficient Diffusion LLMs
Flash-dLLM speeds diffusion LLM inference with I/O-aware KV caching and parallel draft-and-verify decoding.
Flash-dLLM: IO-Aware KV Caching and Parallel Decoding for Fast, Memory-Efficient Diffusion LLMs
Flash-dLLM accelerates diffusion LLM inference with I/O-aware KV caching, up to 11 times faster than Elastic-Cache.
RetroThinker: Enabling Retrospective Thinking in Speech LLMs
RetroThinker is a post-training framework letting the Moshi speech LLM self-correct reasoning mid-stream, adding 11% GSM8K accuracy at similar latency.
NCP-ArchPreview Technical Report: Moving towards Latent Space Language Models through Next Concept Prediction
An 8.9B-parameter latent-space language model using next-concept prediction matches OLMo-3-7B pretraining loss with only 51.3% of the training tokens.