![Integration of the multi-head QNN block into Qwen3-30B-A3B, a 48-layer mixture-of-experts (MoE) model with hidden dimension D = 2048. Left: a standard transformer block, consisting of self-attention preceded by RMSNorm [39] and a sparse MoE feed-forward network that routes each token to 8 of 128 experts, each wrapped in a residual connection. Right: the multi-head QNN block that replaces it at the target layer. The activation x is mixed by the constrained linear transform M1 = Λ2HΛ1, bounded by a tanh nonlinearity, distributed across k = 32 independent 8-qubit QNN heads acting on disjoint 64-dimensional subspaces, and recombined by a second constrained transform M2. The block output is scaled by α and returned to the residual stream. All remaining blocks of the base model are left untouched.](https://pbs.twimg.com/media/HXVTEWcXIAAQeWM.jpg?name=small&format=webp)
ALT Integration of the multi-head QNN block into Qwen3-30B-A3B, a 48-layer mixture-of-experts (MoE) model with hidden dimension D = 2048. Left: a standard transformer block, consisting of self-attention preceded by RMSNorm [39] and a sparse MoE feed-forward network that routes each token to 8 of 128 experts, each wrapped in a residual connection. Right: the multi-head QNN block that replaces it at the target layer. The activation x is mixed by the constrained linear transform M1 = Λ2HΛ1, bounded by a tanh nonlinearity, distributed across k = 32 independent 8-qubit QNN heads acting on disjoint 64-dimensional subspaces, and recombined by a second constrained transform M2. The block output is scaled by α and returned to the residual stream. All remaining blocks of the base model are left untouched.

ALT Four world maps show percentile ranks of modified adjusted net savings per capita in 1990, 2000, 2010 and 2021. Colors range from red for lowest ranks to blue-purple for highest. The United States shifts from green to red-orange, while China shifts from yellow to blue. Canada and Australia remain relatively high-ranked. Gray indicates uncolored areas.

ALT First page of Paul Litvak’s paper, “How Well Does AI Peer Review Work? A Benchmark Built from 100 Planted Errors.” The abstract reports that the best system found 71 errors, the worst found 30, and pooled outputs found 93. All seven errors missed by every system were omissions. False positives were not measured.

ALT Diagram of silent-speech decoding. Dots mark sensor locations on a face and neck beside an EMG waveform. Target text reads “henderson he called you saw that shooting star last night.” The EMG prediction substitutes “could” for “called” and “shouting” for “shooting”; parallel voiced-audio dictation substitutes “call” and “shouting.” Differences are highlighted in red.

ALT Diagram showing bidirectional communication between gaming environments, a gaming agent, and a vision-language model (VLM). Game data includes observations, actions, rewards, and lives. The agent combines recent history, critical history, and full trajectory with game rules, output and format instructions, and textual feedback. A termination module specifies fixed-horizon evaluation.

ALT Screenshot of a paper titled “Can Open Be Too Open? On the Security and Trustworthiness of Fully Transparent LLMs,” by Sheila Zingg, Florian Tramèr, Martin Vechev, and Srdjan Čapkun of ETH Zurich. Below, two columns of text discuss transparency, reproducibility, trust, and security risks in LLMs.






