Dev RL-tunes Qwen 27B against slop with Meta's XAR method, lands top-4 on EQbench
Mysterious_Hearing14 · reddit · 2026-10-16
Developer AlexWortega released Qwen3.8-27B-antislop: a Qwen 27B fine-tune RL'd specifically against "AI slop" stylistic patterns, using an approach in the style of Meta's RL XAR. The result reads without the typical AI flavor and ranks top-4 on EQbench.
Weights are open:
- BF16+MXFP4: HuggingFace AlexWortega/Qwen3.8-27B-antislop
- GGUF: HuggingFace rawmodels/Qwen3.8-27B-antislop-GGUF
A notable community experiment targeting de-sloppification directly as the RL objective — relevant for anyone working on model voice and de-templated output.
More from Models
- Agent Arena: OpenAI and Anthropic Lead on User Enjoyment as Labs Converge on Reliability — arena · 2026-10-17
- Google AI Studio Build gains Max and Plan modes as Gemini 4 Argon surfaces in UI — testingcatalog · 2026-10-17
- NEURA launches Neuron AI, a multimodal PLA model giving robots touch, force feedback and memory — CyberRobooo · 2026-10-17
- Sentience claims per-user models: train, continually update and serve one model per person — soleio · 2026-10-17
- Four Gemini 4 Argon agents beat Minecraft in 27 hours, writing 2,783 programs — GoogleDeepMind · 2026-10-17
- Meta open-sources Muse Spark Open to mixed reception: 'trash' license, nerfed benchmark scores — gabriberton · 2026-10-17