Oct 24, 2026
A subgoal memory graph helps GUI agents reuse meaningful pieces of experience
The method stores reusable subgoal experiences between the granularity of single actions and complete task histories.
Explore Franklin AI
AI Research Papers
Understand new AI research with readable explanations and links to the original papers.
Oct 24, 2026
ValuSpec filters speculative decoding branches before expensive verification
A candidate-valuation module aims to reduce target-model work while preserving greedy decoding output.
Oct 24, 2026
A Petri-net harness studies the boundary between agent reasoning and tool use
Formal supervisory control models capabilities, approvals and budgets, while accounting for the cost of refused actions.
Oct 24, 2026
DIAL studies numerical control over how strongly a model follows an instruction
A lightweight steering adapter aims to vary instruction adherence while preserving fluent responses.
Oct 24, 2026
Adaptive Skill Granularity gives coding agents a route out of failed macros
Runtime routing chooses between primitive actions and larger skills, with explicit recovery when a macro does not fit.
Oct 24, 2026
Muffin uses shared memory instead of synchronous multi-agent debate
An asynchronous framework routes difficult questions to agents that exchange intermediate findings through a shared buffer.
Oct 24, 2026
CMAS tests when an agent system should replace its own architecture
Conservative Meta-Agent Search evaluates architecture changes with replay, risk checks, a simulated canary and rollback.
Oct 24, 2026
FairFRESH treats household food recommendations as a shared-resource problem
A research framework combines preference learning, fair allocation and explicit dietary-rule filtering for households.
Oct 24, 2026
PhysTrace locates when generated video violates physical expectations
A training-free evaluation method compares history-conditioned expectations with observed video windows.
Oct 24, 2026
RIDAR checks research-agent trajectories for integrity violations
A framework defines research-integrity failures, detects them after a run and intervenes during execution.
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.