aiqbench analysis
The paper applies reasoning-conditioned LLM detection to propaganda classification, using population-level refinement and performance-guided resampling of chain-of-thought rationales, without extra explanation annotations. Its reported results are gains over baselines, reaching 85-88% of a supervised fine-tuning upper bound, with interpretability as a stated benefit. For ML readers, the main question is whether the approach holds up beyond these two benchmarks and how much the interpretability claim depends on the quality of the evolved rationales.
Reasoning for Propaganda Detection via Label-Guided CoT Optimisation
arXiv:2610.17271v1 Announce Type: new Abstract: Propaganda detection systems usually operate as black boxes, lacking explanations. To address this, we hypothesise that persuasive techniques contain definitional characteristics with respect to their social-political setting. We introduce ReCoT, a reasoning-conditioned detection framework that captures definitional cues of persuasive techniques via population-level refinement and performance-guided resampling, without requiring additional explanation annotations. On INDI-PROPspan, ReCoT achieves a macro F1_LCS of 0.364, surpassing baselines and reaching 88% of the SFT upper bound on Avg F1_LCS. Similarly, on PTC, ReCoT reaches 85% of the SFT upper bound on Avg F1_LCS, closing much of this gap while remaining interpretable. Importantly, the analysis of the reasoning pool's convergence and update pattern for ReCoT shows effective refinement of reasoning CoTs.
Abstract:Propaganda detection systems usually operate as black boxes, lacking explanations. To address this, we hypothesise that persuasive techniques contain definitional characteristics with respect to their social-political setting. We introduce ReCoT, a reasoning-conditioned detection framework that captures definitional cues of persuasive techniques via population-level refinement and performance-guided resampling, without requiring additional explanation annotations. On INDI-PROPspan, ReCoT achieves a macro F1_LCS of 0.364, surpassing baselines and reaching 88% of the SFT upper bound on Avg F1_LCS. Similarly, on PTC, ReCoT reaches 85% of the SFT upper bound on Avg F1_LCS, closing much of this gap while remaining interpretable. Importantly, the analysis of the reasoning pool's convergence and update pattern for ReCoT shows effective refinement of reasoning CoTs.
Artificial Intelligence (cs.AI)
arXiv:2610.17271 [cs.AI]
(or arXiv:2610.17271v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2610.17271 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 209 words
Related coverage
Links #3: 2026/06 Part 1
LessWrong · Jun 15, 2026
Research Note: Filtering Subversion-Relevant Information From Pretraining Data Is Feasible
LessWrong · Oct 5, 2026
Biological Superintelligence
LessWrong · Jul 30, 2026
Inside D-Wave’s Dual-Rail Qubit Roadmap and the New Haven Bet on Logical Qubits
Quantum Zeitgeist · Oct 11, 2026