• Agent Memory Forgets the Wrong Things, and the Root Cause Is Architecture, Not Prompts. MIT swept 13 configurations and found that where the LLM sits in the memory pipeline — near recall or near rewrite — locks in which class of forgetting errors the system makes. Only a mutation-side hook lifts everything to 91.7–93.2%.

• A Model's "Honesty" Is Inherited, Decided the Moment You Pick the Base. The truthful heads that keep a model grounded in context pass untouched to downstream variants of the same base, even through multimodal conversion. A model's anti-hallucination floor is set at selection time.

• Big Models Read Images Well but Are Nearly Blind to Pixel-Level Tells. MLLM semantic understanding concentrates in the early-to-middle layers, and fine-tuning directly on forensic signals damages those representations. Forgery detection needs a dedicated path to "see noise."

• A 1.5B Model Forecasts Real-World Events and Beats Claude Sonnet 3.5. Oxford turned event forecasting into an RL-trainable objective with tool calls plus GRPO, and the small model edged out the large one on cross-entropy over the same dataset.

Also Notable

• Putting the Data Manifold Under a Microscope to Test Whether Generalization Theory's Geometry Holds Up — Assumptions like intrinsic dimension and curvature, often taken for granted, may not survive contact with real data. Worth a look for anyone reconciling theory with practice.
• Flow Matching Treats Signals as Points in Euclidean Space, Missing the Topology of Data Like fMRI Brain Maps — This adds a topological dimension to the generative framework, relevant if you work on structured signal generation.

Read the full edition →