Tencent's LatentRec replaces inference-time CoT with training supervision, 18.3x lower latency
_reachsumit · x · 2026-11-06
Tencent proposes LatentRec, showing inference-time reasoning artifacts in LLM-based recommenders can be replaced by training-time supervision. A 2.5M-parameter Continuous Thought Module iterates latent reasoning steps over hidden states, Soft Token Selection bridges a 14x norm gap, and Per-step Reasoning Loss supervises intermediate states. On three Amazon benchmarks it improves HR@10 by 4.0–5.0% relative over a CoT-distilled baseline at 18.3x lower latency (batch=1), and the module can be discarded at deployment.
More from Research
- UWaterloo 3D prints Invar36 flexures with directly mounted interferometer optics — jwt0625 · 2026-11-06
- AMD paper: a single generalized MLP upscales textures on-the-fly at render time, saving VRAM and disk — ssh4net · 2026-11-06
- OpenAI researcher invokes 2019 double descent paper in scaling-law spat — willdepue · 2026-11-06
- Encord debuts REACH egocentric hand-tracking model, cutting errors up to 70% — Scobleizer · 2026-11-06
- Only 16% of AI benchmarks measure organizational difficulties of real jobs — ben_burtenshaw · 2026-11-06
- Skillware framework claims 97.3% cost cut as 3B-8B local models beat frontier LLMs — RossPeili · 2026-11-06