Tag

Qwen3-4B

Every Qwen3-4B story we've curated in Bowl of Data, newest issue first — part of our weekly digest across AI, security, blockchain, and engineering.

3 items · 2 issues

Beats AI & ML

Often covered with Large Language Models (LLMs) 2 Qwen3-1.7B 2 vLLM 1

Week 36 · 2026

TTPO: Test-Time Policy Optimization
The paper introduces TTPO, a method designed to optimize LLM reasoning during test-time training by utilizing asymmetric learning signals from pseudo-labels. It effectively mitigates the risks of noisy majority-vote labels by distilling agreeing rollouts and penalizing disagreeing ones.

AI & ML arXiv Source ↗
Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning
This technical report examines the efficacy of using Random Attention for KV cache eviction to optimize LLM inference throughput. The study shows that this method significantly boosts tokens per second in high-concurrency scenarios without sacrificing model accuracy.

AI & ML HuggingFace Papers Source ↗

TTPO: Test-Time Policy Optimization
The paper introduces TTPO, a method designed to optimize LLM reasoning during test-time training by utilizing asymmetric learning signals from pseudo-labels. It effectively mitigates the risks of noisy majority-vote labels by distilling agreeing rollouts and penalizing disagreeing ones.

AI & ML arXiv Source ↗