Oct 22, 2026
AnomalyAgent uses detector disagreement to choose a visual fusion strategy
A post-training framework aligns detector outputs and asks a constrained vision-language model to select from valid fusion actions on high-conflict samples.
Explore Franklin AI
AI Research Papers
Understand new AI research with readable explanations and links to the original papers.
Oct 22, 2026
ToM2Vec learns continuous partner-state representations for multi-agent tasks
The method trains latent reasoning tokens using partner thoughts, then tests coordination across collaborative and competitive environments.
Oct 22, 2026
RTL repair scores can change meaning when reset semantics change
A benchmark audit shows why hardware-repair evaluations need explicit scoring semantics and a specification-only baseline.
Oct 22, 2026
OptScientist searches for optimization laws behind observed data
The framework combines symbolic optimization discovery with an agentic research process, reporting recovery results on 99 physics-derived problems.
Oct 22, 2026
Adaptive multi-token prediction learns how far to draft before verification
AMTP adjusts the speculation horizon online, keeping the exact verifier unchanged and reporting context-dependent decoding gains.
Oct 22, 2026
AI debates can express many values while organizing them differently
A diagnostic framework separates the values an argument invokes from how those values persist and shift across turns.
Oct 22, 2026
Behavioral Localization compresses agent skills around the steps that matter
A SpreadsheetBench study selects influential trajectory positions to train short soft prompts, reporting gains over full-trajectory supervision.
Oct 22, 2026
Evolving semantic indexes improves retrieval when the proxy stays reliable
An evolutionary search study reports gains from changing semantic identifiers, while finding that its small-model proxy does not transfer equally across every component.
Oct 22, 2026
HGM-Dual uses failure history to guide planning-agent harness changes
The framework records edit outcomes and targets constraint violations, reporting more efficient search than open-ended harness evolution in its experiments.
Oct 22, 2026
CoRAL pairs parallel diffusion planning with an autoregressive executor
A hybrid reasoning architecture divides candidate-plan generation from final synthesis, with reported results on vision-language benchmarks.
Good to know
How do I use these AI research explainers?
Start with the overview to understand the paper’s main idea, then follow its original paper link for the methods, results and limitations.
Can I suggest a research paper?
Use Submit Research to send the original paper URL and explain why it is useful. An editor reviews the suggestion before publication.