AIResearch
Reflexion Framework: Verbal Self-Critique Loops for Language Agents
{{Reflexion}} (Shinn et al. 2023) is a framework where a language model agent generates verbal self-critique after each attempt and uses that critique to improve subsequent attempts, without any parameter updates.
FID on Training-Distribution Data: The Tautological Benchmark Win
{{Frechet Inception Distance}} (FID) measures how closely a model's outputs match a reference distribution. When that reference distribution is the model's own training data, winning on FID is tautological and reveals nothing about generalization.
LoRA Cross-Model Adapter Pattern: Wiring Pretrained Models Together
A growing research pattern uses small {{LoRA}}-style adapters to bridge two large pretrained models rather than train a unified architecture from scratch. The pattern is legitimate engineering but often gets oversold as a new "framework" when the heavy lifting is done by the frozen base models.
UniMesh: Unified 3D Mesh Framework or LoRA Adapter Between BAGEL and Hunyuan3D?
UniMesh (arXiv 2604.17472) markets itself as the first unified framework for 3D generation and understanding, but its actual training contribution is a single small {{LoRA}} adapter wiring BAGEL's image latent space to Hunyuan3D's conditioner. Benchmarks selectively omit closed-source SOTA and win only on FID against training-distribution data.