From the Archive

From the Archive — 2026-06-22

2026-06-22

A short reading list of papers that landed this week. Each one sits at the edge of a longer conversation already underway in the research literature.

Retrieval Collapses When AI Pollutes the Web

Hongyeon Yu, Dongchan Kim, Young-Bum Kim · arXiv:2602.16136

As AI-generated content saturates the web, a feedback loop emerges where retrieval systems designed to surface evidence increasingly feed on their own synthetic outputs, creating what this work frames as a structural collapse rather than a simple quality problem. The risk cuts deeper than contamination alone—even when answer accuracy holds steady, the underlying evidence base quietly shifts toward synthetic sources, a kind of epistemic drift that recursive training on synthetic data has already been shown to amplify across generations. Yet the paper also surfaces an asymmetry: LLM-based rankers suppress adversarial content more effectively than traditional retrieval baselines, suggesting the problem may not be retrieval itself but rather the absence of verification mechanisms, which raises the question of whether systems like gated feedback loops in RAG pipelines might interrupt the self-reinforcing cycle—or whether any closed-loop learning in a contaminated ecosystem remains fundamentally fragile.

Go deeper into Language, Text, and Discourse

Retrieval-Augmented Generation StrategiesTraining Dynamics And GeneralizationReasoning Trace ReliabilityAI Text Perception And Authorship
Explore these inquiring lines →

The Ghost Couple: Correlated LLM Name Priors and Their Haunting of the Web and Academic Publishing

Michał Brzozowski, Neo Christopher Chung · arXiv:2606.02184

The discovery of these correlated phantom authors cuts to the heart of a deeper question about what LLMs actually encode: models don't just converge on plausible outputs, they converge on predictable co-occurrence patterns, creating signatures as legible as fingerprints. This matters not because ghost names are merely a curiosity, but because they've begun polluting real scholarly infrastructure—systems designed to catalog genuine knowledge now index synthetic ensembles alongside real research, blurring the boundary between fabrication and documentation. The deeper tension is that LLM outputs represent subjective belief distributions encoded in training data, not objective observation—and when those distributions leak into citation graphs and DOI registries, we lose the ability to distinguish signal from artifact. How do we design scholarly gatekeeping systems when the generative models themselves encode brand-specific hallucinations that downstream tools cannot distinguish from legitimate publication metadata?

Go deeper into Language, Text, and Discourse

Scaling, Sparsity & Data Trade-offsAI Text Perception And AuthorshipLLM Cognitive LimitationsTraining Dynamics And Generalization
Explore these inquiring lines →

An Enigma of Artificial Reason: Investigating the Production-Evaluation Gap in Large Reasoning Models

Mingzhong Sun, Teresa Yeo, Armando Solar-Lezama, et al. · arXiv:2606.01462

Recent work has documented a curious asymmetry in how humans and large language models handle reasoning: whether reasoning steps actually causally shape outputs or merely rationalize predetermined conclusions remains contested, and this paper sharpens that question by isolating a production-evaluation gap unique to machine reasoning. While humans prove nearly as competent at grading flawed math solutions as solving them, frontier LRMs show dramatic collapse in evaluation—near-perfect generation but 48% accuracy on invalid reasoning with correct answers—suggesting a systemic bias toward confirmation rather than verification. The paper's finding that answer validity drives model verdicts rather than step-by-step scrutiny points to a deeper training-induced pathology: models optimized to produce chains toward correct answers may never learn to robustly reject invalid reasoning paths that happen to land on right results. This raises a pressing question about what training regimes could cultivate genuine evaluative reasoning rather than answer-chasing, and whether post-hoc RL interventions could decouple reasoning verification from answer confirmation.

Go deeper into Reasoning, Retrieval, and Evaluation

Chain-of-Thought FaithfulnessReasoning Model Quality & TrainingReasoning Model Failure ModesReasoning Trace Reliability
Explore these inquiring lines →

All briefs →