AIBullisharXiv – CS AI · Jun 117/10
🧠Researchers introduce ICALens, a new method for interpreting language model representations using independent component analysis (ICA) instead of expensive sparse autoencoders (SAEs). The approach efficiently recovers interpretable directions without requiring large neural dictionary training, achieving competitive performance on standard benchmarks while offering a faster, more accessible alternative for LLM analysis.
AINeutralarXiv – CS AI · Jun 57/10
🧠Researchers present a three-step methodology for identifying and validating attention-head circuits in transformer models using spectral analysis, pattern filtering, and causal ablation. The technique successfully isolates core computational circuits across multiple model sizes and architectures without requiring labeled data or gradient attribution.
AINeutralarXiv – CS AI · Feb 277/106
🧠Researchers discovered that a Qwen 32B AI model can detect when concepts have been injected into its context, even though it denies this capability in its outputs. The introspection ability becomes dramatically stronger (0.3% to 39.9% sensitivity) when the model is given accurate information about AI introspection mechanisms.
AINeutralarXiv – CS AI · Jun 236/10
🧠Researchers have developed a framework using Sparse Autoencoders to extract and interpret visual, textual, and multimodal concepts from Vision Language Models, achieving 45% improvement in visual concept quality compared to existing methods. This advancement provides structured insights into how VLMs process joint image-text information, addressing a critical gap in AI interpretability research.
AINeutralarXiv – CS AI · Jun 236/10
🧠Researchers present novel a-priori generalization bounds for nearly-linear neural networks that do not require training to evaluate. This represents a theoretical breakthrough in understanding how well neural networks generalize to unseen data, with bounds that become non-vacuous specifically for networks operating close to linear regimes.
AINeutralarXiv – CS AI · Jun 236/10
🧠Researchers introduce INFORM, an interpretability framework for analyzing multi-expert LLM orchestration systems, revealing that frequently routed experts often serve as structural hubs with minimal functional impact while sparsely selected experts can be critically important. The study challenges conventional assumptions about expert importance in collaborative AI systems and provides tools for understanding opaque decision-making in complex model architectures.
AINeutralarXiv – CS AI · Jun 96/10
🧠Researchers introduce Contribution Weights, a new metric for analyzing transformer attention that accounts for value vector geometry alongside attention weights. The approach more accurately identifies semantically critical tokens than traditional attention-based metrics and reveals that attention sinks actively suppress information rather than passively storing excess attention.
AINeutralarXiv – CS AI · Jun 26/10
🧠Researchers introduce Partial Information Decomposition (PID), a framework for analyzing how multimodal language models integrate vision and language inputs by separating unique, redundant, and synergistic contributions. The analysis reveals distinct modality-use patterns across task types and identifies visual dominance as a bottleneck in audio-visual fusion systems.
AINeutralarXiv – CS AI · Jun 26/10
🧠Researchers propose 'Model Science,' a systematic discipline for understanding AI models beyond traditional benchmarking. The framework consolidates analysis around four functional perspectives—Verify, Explore, Steer, and Refine—and emphasizes deep study of individual models rather than population-level comparisons, drawing lessons from established sciences like neuroscience and medicine.
AINeutralarXiv – CS AI · Jun 26/10
🧠Researchers analyzed how language models make decisions by tracing answer scores across neural network layers in 9,000 MMLU trajectories, finding that correct answers are often unstable and that attention mechanisms better preserve correctness than MLP layers. The study reveals decision-making is a distributed process rather than a final-layer phenomenon, with implications for understanding model reliability and interpretability.
🧠 Llama
AINeutralarXiv – CS AI · May 296/10
🧠Researchers introduced ReasonOps, an unsupervised method for analyzing chain-of-thought traces from large language models that identifies seven universal reasoning operators (backtracking, inferring, hypothesizing, etc.) appearing consistently across 12 different LLM families. The framework enables model identification, correctness prediction, and early quality estimation without manual annotation, revealing that each model family has a distinctive reasoning fingerprint.
AINeutralarXiv – CS AI · May 286/10
🧠Researchers studying DeepSeek-V3 discovered that Large Language Models encode syntactic and semantic information in mathematically separable, linear patterns within their hidden layers. By averaging representations of sentences with shared structure or meaning, they created 'centroids' that capture significant linguistic information, revealing that syntax and semantics are processed through distinct, partially decoupled mechanisms across different layers.
AINeutralarXiv – CS AI · May 276/10
🧠Researchers introduce CUDAnalyst, a new analysis framework that reveals how large language models make planning decisions when generating CUDA kernels by decomposing feedback signals. The study demonstrates that explicit planning helps only when feedback is well-aligned and that effective planning emerges from structured multi-feedback interactions, with findings showing robustness across different models and workloads.
AINeutralarXiv – CS AI · May 126/10
🧠Researchers analyzed how Qwen3-VL-8B, a multimodal transformer, encodes visual interestingness—a measure derived from human engagement data—without explicit supervision. Using neuroscience-inspired methods, they found that the model's internal representations align with human-derived interestingness scores, suggesting transformers may capture principles of human attention and perception.
AINeutralarXiv – CS AI · May 96/10
🧠Researchers have developed a visual fingerprinting method to compare Large Language Model outputs across different generation conditions by analyzing linguistic choices in content, expression, and structure. This approach enables pattern recognition in LLM behavior that is difficult to detect through individual responses or standard metrics, advancing model evaluation and prompt optimization techniques.
AINeutralarXiv – CS AI · Apr 106/10
🧠Researchers introduce Step-Saliency, a diagnostic tool that reveals how large reasoning models fail during multi-step reasoning tasks by identifying two critical information-flow breakdowns: shallow layers that ignore context and deep layers that lose focus on reasoning. They propose StepFlow, a test-time intervention that repairs these flows and improves model accuracy without retraining.
AIBullisharXiv – CS AI · Mar 36/106
🧠Researchers introduce CIRCUS, a new method for discovering mechanistic circuits in AI models that addresses uncertainty and brittleness issues in current approaches. The technique creates ensemble attribution graphs and extracts consensus circuits that are 40x smaller while maintaining explanatory power, validated on Gemma-2-2B and Llama-3.2-1B models.
