https://peertube.watch/videos/watch/b0386b42-a420-4cb5-b47b-ff785506238c Mr. Peace is an AI-made character; voice is AI-generated; script written with AI assistance. AI-generated presenter. No affiliate links in this video. This is a task-specific comparison. The Qwen3.5-4B model was trained (supervised + reinforcement fine-tuning, the SLIC recipe) for this exact job with simulator feedback; DeepSeek-V4.1-Flash Think (763B) was prompted, not trained for it. The result concerns four-specification amplifier sizing on one analog circuit optimization benchmark, the researchers' own, not general AI capability. The paper was first posted on arXiv on November 10, 2026; the authors note it was accepted at NeurIPS 2026. In a NeurIPS 2026 paper by Xu et al. (SLIC), a reinforcement-fine-tuned Qwen3.5-4B passed 87.1% of 840 in-distribution analog-circuit tasks within 8 simulator turns (4-spec Pass@1, PTM 65 nm), vs 38.6% for the 763B DeepSeek-V4.1-Flash in Think mode (Table 1, mean of 3 runs). The catch: a bigger model still won. GPT-6-Astra passed 90.6% (one run). The untuned Qwen3.5-4B passed 2.3%, and DeepSeek-V4.1-Flash generated the trajectories used for the small model's supervised training. On held-out circuit topologies the 4B model passed 70.6% vs 26.7% for DeepSeek and 86.2% for GPT-6-Astra (Table 2). Source: Xu, Gao, Ko, Hong, Lu, Yu, Zhang, Boning, Han, 'SLIC: Reinforcement Fine-Tuning Small LMs for Multi-Turn Analog Circuit Optimization', arXiv:2611.13452v1 - arXiv PREPRINT copy (Nov 10, 2026); arXiv comment: "Accepted at NeurIPS 2026." https://arxiv.org/abs/2611.13452 . Numbers from Tables 1 and 2. Names are trademarks of their respective owners; Peace Call is not affiliated with or endorsed by them. Charts drawn by code from the source's numbers. Paper: https://arxiv.org/abs/2611.13452 (version v1, posted Nov 10, 2026). AI Disclosure: Mr. Peace is a fictional AI character using a synthetic voice. This video is AI-assisted and does not depict a real person speaking. Research Disclosure: Accepted at NeurIPS 2026 (per the authors' arXiv note); the arXiv version is the preprint copy. Results apply to the reported benchmark and experimental conditions, not overall model superiority. Trademark Disclosure: Names are trademarks of their respective owners. Peace Call is not affiliated with or endorsed by them. More from Peace Call: https://getpeacecall.com/?utm_source=peertube Original on YouTube: https://youtube.com/shorts/aUb_qtjFhMI #AI #SmallVsBig Sun, 15 Nov 2026 13:57:48 GMT https://validator.w3.org/feed/docs/rss2.html PeerTube - https://peertube.watch https://peertube.watch/lazy-static/avatars/3b284627-b75f-409e-ac2e-10f4c2946f43.png https://peertube.watch/videos/watch/b0386b42-a420-4cb5-b47b-ff785506238c All rights reserved, unless otherwise specified in the terms specified at https://peertube.watch/about and potential licenses granted by each content's rightholder.