Fetching from the wire…
01
02
03
04
05
06
07
08
Research2026-09-21 · source-backed
A round-trip test has a generator turn a procedurally-generated arithmetic expression into a word problem and a separate extractor recover the expression from the prose alone, with symbolic equivalence as an exact oracle and no judge in the loop. All pairwise combinations of sixteen models produce a communication matrix whose marginals separate generation quality from extraction quality, and results change when you swap which model generates and which extracts. If you pass free-text intermediates between agents, here's a cheap oracle-backed way to measure what your handoff drops.
Each link below shares sources, entities, or timing with this story.
Li, Huo, and Johnson show that one-way message flow between agents produces neither mimicry nor solo behavior but an entirely novel dynamical state, at identical temperature settings. It's conceptual rather than quantitative, but the implication for orchestrator-worker fan-out...
One number predicts whether your agent finishes the task, and it isn't the benchmark score. Shubhra Mittal's paper (arXiv 2609.01660) analyzed 10,664 trajectories across nine models spanning 1.2B to 671B parameters and found task success follows P(n) = p^n, where p is a single...
APort Vault replays 4,371 human-written attacks from a public CTF against a live payment agent, across 14 models from 8 labs, five policy configurations and two tracks, for 225,964 total evaluations. At Levels 2 through 4, transfers to recipients the passport didn't permit num...
arXiv 2609.09553 shows cipher-based covert-communication jailbreaks no longer need fine-tuning on an encrypted corpus. In-context learning is enough, and alignment is significantly weakened or bypassed once the exchange runs through the learned encoding. Demonstrated against m...
Eight teams per setting formed independently from one base model, each agent keeping a private notebook across ten formation episodes, then role-matched agents were traded between teams (arXiv 2609.05279). Against a placebo reproducing roster-change disruption without changing...
CAFE (arXiv 2608.24794) makes corrective feedback an in-trajectory intervention the agent chooses to request, using one shared-parameter model alternating between search-agent and critic roles. Online RL shapes request returns from a prompt-level call-versus-skip success gap;...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.