qwen3-6-35b-a3b-6a005d48·2 events·first seen Aliases: Qwen3.6-35B-A3B
Researchers use learned soft prefixes (opaque continuous vectors) to probe the logical stability of LLMs on syllogistic reasoning benchmarks without modifying model weights. Across Qwen3.6-35B-A3B MoE, Qwen3-8B, and Gemma 4 31B, successful prefixes redirect correct answers at flip rates of 72–90% for Qwen3.6 MoE and 54–56% for Gemma, far exceeding random controls by 37–99 percentage points. The dominant effect is a broad answer-preference bias rather than symbol-level forcing, and the bias generalizes across unseen logical forms and prompt interfaces. Model-specific differences in how this bias manifests suggest substantial variation in logical robustness across architectures.
Qwen published Qwen3.6-35B-A3B, a 35B-parameter mixture-of-experts image-text-to-text model with 3B active parameters, on Hugging Face. The model supports conversational use and is compatible with Azure deployment endpoints. With over 5.9 million downloads and 2,000 likes, it has seen substantial community uptake.