arXiv subjects

Santosh Premi Adhikari

Publications and source records attributed to Santosh Premi Adhikari.

3 recordsLinked to original sources

arXiv2026

Doom Researching: A Conceptual Framework for Repetitive AI-Assisted Information Seeking, Cognitive Offloading, and the Illusion of Knowing

Generative artificial intelligence (GenAI) systems such as ChatGPT, Claude, and Gemini have made information seeking faster, more conversational, and more cognitively comfortable. These affordances can support learning and productivity, but they can also encourage a repetitive pattern in which users continue querying AI systems for explanations, summaries, comparisons, plans, and reassurance without converting those interactions into durable understanding, decisions, or finished work. This conceptual paper proposes the term doom researching to describe this AI-mediated pattern of repetitive information seeking without proportional synthesis or output. Building on research on doomscrolling, information seeking, cognitive offloading, transactive memory, human-AI interaction, productivity loss, and the illusion of knowing, the paper develops a framework in which fluent AI responses reduce cognitive effort, inflate perceived knowledge, and increase the likelihood of further querying. The framework distinguishes doom researching from doomscrolling, cyberchondria, ordinary research, and productive AI-assisted work. It introduces a formal model of the doom researching loop, a candidate risk index for empirical measurement, and testable propositions for future studies. It then extends the construct through the lens of the extended mind thesis, distinguishing assistive, substitutive, and disruptive forms of cognitive offloading, and connects individual doom researching to the broader literature on AI-driven homogenization and knowledge collapse. The paper argues that doom researching is not simply "too much AI use" but a misalignment among inquiry, metacognition, and output. The goal is to provide a vocabulary and research agenda for studying when AI-assisted inquiry becomes a substitute for thinking, synthesis, and action.

cs.HC↗

arXiv2026

Convergence Theory for Iterative LLM-Based Neural Architecture Search: A Parametric Cross-Entropy Framework with Closed-Form Proxy Reliability

Large language models (LLMs) are increasingly used as generators in iterative neural architecture search (NAS), yet no formal convergence theory exists for this class of algorithms. We model iterative LLM-NAS as a parametric Cross-Entropy (CE) method over executable programs and prove six results: (1) iterative LLM fine-tuning on elite architectures is equivalent to the CE update restricted to the LLM parametric family; (2) expected architecture quality is monotonically non-decreasing across cycles; (3) elite-set probability converges to a fixed point at a geometric rate C_t >= 1-(1-rho_0)^t; (4) delta-based generation achieves a strictly higher valid-generation rate than full-code generation under a first-order Markov token-error model; (5) the MinHash-Jaccard novelty filter prevents mode collapse; (6) proxy reliability admits the closed-form rho_S = (6/pi) arcsin(rho_P(SNR)/2), yielding the practical diagnostic sigma^2_arch >> sigma^2_noise as a necessary condition for trustworthy proxy-based rankings. Testing against a 22-cycle, three-LLM, six-dataset experiment with 3,300 generated architectures confirms two predictions quantitatively, two at direction-of-effect level, and explains the proxy-reliability ceiling effect previously reported empirically but left unexplained.

cs.LG↗

arXiv2026

Delta-Based Neural Architecture Search: LLM Fine-Tuning via Code Diffs

Large language models (LLMs) show strong potential for neural architecture generation, yet existing approaches produce complete model implementations from scratch -- computationally expensive and yielding verbose code. We propose Delta-Code Generation, where fine-tuned LLMs generate compact unified diffs (deltas) to refine baseline architectures rather than synthesizing entire models. Our pipeline iteratively fine-tunes the LLM via LoRA on curated architectures from the LEMUR dataset, with MinHash-Jaccard novelty filtering for structural diversity. We evaluate three 7B-class LLMs -- DeepSeek-Coder-7B, Qwen2.5-Coder-7B, and Mistral-7B -- across six datasets (CIFAR-10, CIFAR-100, MNIST, SVHN, ImageNette, CelebA) using a 22-cycle protocol (1,100 candidates per LLM). All three substantially surpass the full-generation baseline (50.6% valid rate, 42.3% mean first-epoch accuracy): DeepSeek-Coder reaches 75.3% valid rate and 65.8% mean accuracy; Qwen2.5-Coder 72.1%/64.6%; Mistral 66.6%/66.1%. On CIFAR-10, best first-epoch accuracies reach 85.5% (Mistral), 85.2% (DeepSeek), 80.6% (Qwen) -- well above 63.98% full generation and 71.5% for the concurrent approach of Gu et al. Output lengths are 30-50 lines versus 200+ for full generation (75-85% reduction). A 50-epoch study confirms the 1-epoch proxy preserves rankings (Mistral: Spearman $ρ$ = 0.926). Delta-based generation is a token-efficient, multi-domain, LLM-agnostic alternative to full-model synthesis for LLM-driven NAS.

cs.LG↗