All issues

design(vlm): support multimodal KV prefix reuse

#1455

· MAN$I VERMA (mansiverma897993) opened

on Sep 11, 2026

QAIRT backend: tools parameter silently dropped in apply_chat_template, tool calling non-functional

#1454

· EURUS (EurusZhang) opened

on Sep 11, 2026

Outdated llama.cpp in geniex doesn't support Spark X2.5 (spark2_5 architecture), fails to run Spark-X2.5-4B-GGUF

#1453

· EURUS (EurusZhang) opened

on Sep 11, 2026

Tool calls not working with MiniCPM5 models: GenieX only parses OpenAI JSON format, not XML-style function calls

#1452

· EURUS (EurusZhang) opened

on Sep 11, 2026

Update README for current model sources and Android release

#1444

· MAN$I VERMA (mansiverma897993) opened

on Sep 10, 2026

Improve QAIRT model performance on Snapdragon X Plus 10-Core machines

#1433

· flatlines opened

on Sep 7, 2026

Android JNI: applyChatTemplate failure is returned as an empty formattedText instead of an error

#1432

· PERIN François (sGeeK44) opened

on Sep 7, 2026

Qualcomm Adreno OpenCL driver (CB.dll) crashes llama_cpp/GGUF serving with 0xC0000409

#1431

· Joshua Timothy (jdtimothy) opened

on Sep 4, 2026

ggml-hexagon on Hexagon v73 (QCM6690): NPU output is byte-identical regardless of the prompt — silent corruption, no crash, `graph splits = 3`

#1429

· PERIN François (sGeeK44) opened

on Sep 4, 2026

Add transaction-safe lineage validation for retained chat caches

#1413

· Sam Kwak (samkwak188) opened

on Aug 31, 2026

Add fidelity-to-teacher KLD and long-context retrieval eval types

#1411

· samuelsalfati opened

on Aug 30, 2026

# Feature request: MoE expert streaming / caching to run large MoE models

#1409

· AHMETBAYIR opened

on Aug 29, 2026