Oct 14, 2026
CarryOn organizes long-running agent context around reusable task objects
A context runtime carries progress forward through evidence-bound records and a live tail of recent interactions.
Explore Franklin AI
AI Research Papers
Understand new AI research with readable explanations and links to the original papers.
Oct 14, 2026
Jeva builds explicit absence and reusable options into decision models
A decision architecture aims to preserve choices across candidate reordering while reusing representations between requests.
Oct 14, 2026
AesCode separates visual style from content correctness in generated code
AesCode combines a design graph, deterministic checks and visual judging to train code generation for editable visual artifacts.
Oct 14, 2026
Scenario-Actor adapts decoding temperature to a communicative setting
A research method uses model hidden states to represent a situation and select a temperature instead of relying only on a fixed setting.
Oct 14, 2026
Arbiter checks code with counterexamples that explain a failed verdict
An agentic verifier is trained to judge programs and produce checkable failing inputs without using official submissions.
Oct 14, 2026
Physical RSI survey asks whether robots improve their own learning process
A survey separates better task execution from verified improvements to the procedure a robot uses to learn.
Oct 14, 2026
MFG trains molecular optimization to use feedback across several turns
A reinforcement-learning framework rewards progressive property improvements while retaining similarity to a starting molecule.
Oct 14, 2026
Harness RSI study separates agent revisions from the rules that admit them
A theoretical framework examines persistent changes to prompts, memory, workflows and code while model weights remain fixed.
Oct 14, 2026
BLADE audits how agents use information when tracing a failure
A training-free framework distinguishes introducing, propagating and repairing errors at handoffs between agents.
Oct 14, 2026
P2T derives token rewards from an existing process reward model
A gradient estimator attributes a sequence-level reward to individual tokens without training another reward model.
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.