Oct 15, 2026
DDRG checks disagreements inside reasoning traces before repairing an answer
Graph alignment targets conflicting intermediate claims while retaining the original sampled reasoning bank.
Explore Franklin AI
AI Research Papers
Understand new AI research with readable explanations and links to the original papers.
Oct 15, 2026
COMPASS estimates model-pair compatibility before expensive training
A two-part diagnostic ranks model pairs for cross-modal alignment and checkpoint merging.
Oct 15, 2026
Structured Recipe Optimization makes post-training choices reusable across experiments
SRO combines an adaptive recipe space with evidence-guided candidate construction and multi-objective trial selection.
Oct 15, 2026
PGLA adds pairwise comparisons to sampled-answer voting
The framework adjusts answer rankings using comparative confidence alongside the number of supporting reasoning paths.
Oct 15, 2026
Training agent confidence against the final outcome improves failure prediction
An experimental calibration method asks whether the whole agent attempt will succeed, rather than whether one step looks correct.
Oct 15, 2026
Correctly reading an agent state does not guarantee the right action
Paired factual probes reveal decision failures after models have already understood the relevant observations.
Oct 15, 2026
Active-inference study separates clearer body signals from stronger preferences
A foraging simulation shows why survival alone cannot explain how an agent prioritizes competing needs.
Oct 15, 2026
AgentStripes separates agent capability scores from evidence of engineering controls
The proposed maturity model grades supporting evidence but has not been validated as a predictor of safety or reliability.
Oct 15, 2026
Agent memory needs to distinguish suggestions from permission
Grounding Awareness tests whether a personal agent remembers what a user actually accepted, revised or limited.
Oct 15, 2026
Historical agent audits need to check model memory as well as retrieved sources
A point-in-time study finds future-fact recall does not translate into a simple, predictable search bias.
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.