Issues

• fix(attention): include cached prefixes in chunked prefill

  inclusionAI/AReno#635 · Filter by author xsuler opened on Oct 26, 2026 · v0.11.0
  

• perf(sampling): add Qrita-style sorting-free top-k/top-p sampling

  inclusionAI/AReno#634 · Filter by author xsuler opened on Oct 26, 2026 · v0.11.0
  

• perf(inference): remove per-token host synchronization from decode metadata

  inclusionAI/AReno#633 · Filter by author xsuler opened on Oct 26, 2026 · v0.11.0
  

• feat(models): add GPT-OSS support

  enhancement
  inclusionAI/AReno#632 · Filter by author xsuler opened on Oct 26, 2026 · v0.11.0
  

• feat: add TPU backend support

  enhancement
  inclusionAI/AReno#631 · Filter by author xsuler opened on Oct 26, 2026 · v0.11.0
  

• feat(jev): support full-parameter training on MLX

  enhancement
  inclusionAI/AReno#630 · Filter by author xsuler opened on Oct 26, 2026 · v0.11.0
  

• feat(jev): support LoRA training on CUDA

  enhancement
  inclusionAI/AReno#629 · Filter by author xsuler opened on Oct 26, 2026 · v0.11.0
  

• feat(jev): support full-parameter training on CUDA

  enhancement
  inclusionAI/AReno#628 · Filter by author xsuler opened on Oct 26, 2026 · v0.11.0
  

• docs: Add a reproducible guide for the RSI demo

  kind/documentation
  inclusionAI/AReno#612 · Filter by author sofan110 opened on Oct 21, 2026
  

• feat: Add a RSI demo

  enhancement
  inclusionAI/AReno#611 · Filter by author sofan110 opened on Oct 21, 2026
  

• feat: Add a composable RSI experimental framework

  enhancement
  inclusionAI/AReno#610 · Filter by author sofan110 opened on Oct 21, 2026
  

• docs: Enrich the Get Started - Welcome page and improve its reading experience

  enhancement
  inclusionAI/AReno#604 · Filter by author adohe opened on Oct 20, 2026 · v0.11.0
  

• 【2026 开源轻训营专属 Issue】GaLore 全参数优化 / GaLore for Full-Parameter Training

  蚂蚁开源轻训营专属轻训营-中高难度
  inclusionAI/AReno#593 · Filter by author xsuler opened on Oct 10, 2026

• 【2026 开源轻训营专属 Issue】多采样共享 Prompt KV / Share Prompt KV across Rollout Samples

  蚂蚁开源轻训营专属轻训营-中高难度
  inclusionAI/AReno#592 · Filter by author xsuler opened on Oct 10, 2026

• 【2026 开源轻训营专属 Issue】自适应激活压缩 / Adaptive Saved-Activation Compression

  蚂蚁开源轻训营专属轻训营-中高难度
  inclusionAI/AReno#591 · Filter by author xsuler opened on Oct 10, 2026

• 【2026 开源轻训营专属 Issue】KIVI 低比特 KV 缓存 / KIVI Low-Bit KV Cache

  蚂蚁开源轻训营专属轻训营-中高难度
  inclusionAI/AReno#590 · Filter by author xsuler opened on Oct 10, 2026

• feat: add opt-in LoRA-FA to reduce adapter activation memory

  inclusionAI/AReno#589 · Filter by author xsuler opened on Oct 10, 2026
  1

  

• 【2026 开源轻训营专属 Issue】CCE 词表投影融合 / Fused Vocabulary Projection with CCE

  蚂蚁开源轻训营专属轻训营-中高难度
  inclusionAI/AReno#588 · Filter by author xsuler opened on Oct 10, 2026

• feat: implement Jet-RL mixed-precision FP8/BF16 training and inference for weights, activations, and gradients

  inclusionAI/AReno#560 · Filter by author xsuler opened on Sep 14, 2026
  

• feat: Support Hybrid Parameter-Efficient RL (Mixed Full-Parameter & LoRA Fine-Tuning)

  inclusionAI/AReno#555 · Filter by author indevn opened on Sep 10, 2026
  

• [RFC] Integrate MTP / speculative decoding into the AReno rollout engine

  inclusionAI/AReno#545 · Filter by author Dayuxiaoshui opened on Sep 2, 2026
  

• Add FP8 KV-cache storage for inference and agentic rollout

  area/enginearea/performancearea/servingkind/featurepriority/backlog
  inclusionAI/AReno#533 · Filter by author xsuler opened on Aug 31, 2026
  

• Track memory-saving quantization for AReno training and rollout

  area/enginearea/performancekind/featurepriority/important-longterm
  inclusionAI/AReno#531 · Filter by author xsuler opened on Aug 31, 2026
  

• [RFC] FP8 / INT4 weight quantization for AReno (CUDA train & decode)

  inclusionAI/AReno#514 · Filter by author CAICAIIs opened on Aug 25, 2026
  2

  

• [bug] /v1/chat/completions returns 400 when stream=true, breaking Agent framework integration

  inclusionAI/AReno#503 · Filter by author tangxinyao opened on Aug 21, 2026