AINeutralarXiv – CS AI · Jun 26/10
🧠Researchers argue that benchmarking vision-language models for urban perception tasks must account for human disagreement and measurement reliability rather than treating consensus as ground truth. A study of seven VLMs evaluated on 100 Montreal street scenes reveals that model performance correlates with inter-annotator reliability, highlighting the need for transparent uncertainty reporting in AI evaluation frameworks.
AINeutralarXiv – CS AI · Jun 26/10
🧠Researchers introduce Reward Partition Optimization (RPO), a new method for training language models that eliminates the need for value function estimation in preference-based learning. RPO simplifies the optimization process by normalizing rewards through partition-based formulations, demonstrating superior performance compared to existing approaches like DRO and KTO across multiple model architectures.
AINeutralarXiv – CS AI · Jun 26/10
🧠Researchers propose Base-Aligned Model Collaboration (BACo), an inference-time framework that dynamically combines base and aligned language models to improve both output diversity and quality simultaneously. The method uses token-level routing strategies based on uncertainty signals, achieving a 21.3% joint improvement in diversity-quality metrics without requiring expensive retraining or multi-pass decoding.
AINeutralarXiv – CS AI · Jun 16/10
🧠Researchers propose Canopy Entropy (CE*), a new metric that reveals fine-tuning reorganizes uncertainty in language models rather than simply reducing it. The measure shows that fine-tuned models convert token-level uncertainty into more semantically meaningful and informative outputs, fundamentally changing how we understand model alignment and information generation.
AINeutralarXiv – CS AI · May 296/10
🧠Researchers introduce Thoughts-as-Planning, a novel framework that optimizes reasoning chains in large language models by modeling them as sequential decision-making processes over a latent semantic space. The method uses learned world models to simulate how edits to reasoning chains affect outputs, enabling efficient planning through gradient descent or reinforcement learning while supporting multi-scale abstraction across token, segment, and instruction levels.
AINeutralarXiv – CS AI · May 296/10
🧠Researchers propose that representation alignment across AI models stems from linear encoding of object-attribute relationships, with quality determined by signal strength, architectural bias, and training noise. The study demonstrates that sparse autoencoders extract these linear features more effectively than dense models, and that data scarcity significantly impacts cross-model alignment in both language and embedding models.
AINeutralarXiv – CS AI · May 286/10
🧠Researchers introduce ORBIT, a reinforcement learning framework that uses dynamically generated rubrics to fine-tune large language models for open-ended medical dialogue tasks. The approach achieves state-of-the-art performance on medical benchmarks with minimal training data, addressing the challenge of applying RL to complex tasks where traditional scalar reward signals are inadequate.
AINeutralarXiv – CS AI · May 276/10
🧠Researchers introduce PICACO, a novel in-context alignment method that optimizes meta-instructions to help large language models better understand and balance multiple, often conflicting human values without fine-tuning. The approach uses total correlation optimization to improve alignment across up to 8 distinct values while reducing noise, addressing a key limitation where LLMs struggle to reconcile competing preferences in single prompts.
AINeutralarXiv – CS AI · May 126/10
🧠Researchers introduce an anchor-projection framework that enables behavioral directions to transfer across different large language model families by mapping their diverse hidden representations into a shared coordinate space. The approach achieves high cross-model alignment (0.83 ten-way detection accuracy) without fine-tuning, demonstrating that interpretability and control mechanisms can be standardized across architecturally different models.
🧠 Llama
AINeutralarXiv – CS AI · May 116/10
🧠Researchers introduce Prune-OPD, a framework that optimizes on-policy distillation for AI reasoning models by detecting when student predictions diverge from teacher guidance and dynamically truncating unreliable training sequences. The method reduces training time by 37-68% on challenging math benchmarks while maintaining or improving performance.
AINeutralarXiv – CS AI · May 96/10
🧠Researchers introduce Annotator Policy Models (APMs), interpretable machine learning models that extract and visualize annotators' implicit safety policies from labeling behavior alone. By revealing disagreement sources—operational failures, policy ambiguity, and value pluralism—APMs enable more transparent and inclusive AI safety policy design without requiring costly additional annotation.
AINeutralarXiv – CS AI · May 96/10
🧠Researchers introduce CrossCult-KIBench, a benchmark dataset for evaluating how multimodal large language models (MLLMs) handle cross-cultural knowledge insertion across English, Chinese, and Arabic contexts. The work reveals that current AI models struggle to adapt to specific cultural contexts without degrading performance in other cultures, establishing a new research direction for culturally-aware AI systems.
AINeutralarXiv – CS AI · May 46/10
🧠Researchers introduce LOCA, a method for identifying why specific jailbreak attacks succeed against safety-trained LLMs by pinpointing minimal, causal changes in intermediate representations. The approach provides local explanations for individual jailbreak instances rather than global theories, achieving refusal induction with an average of six interpretable changes compared to prior methods requiring 20+.
🧠 Llama
AINeutralarXiv – CS AI · Apr 156/10
🧠Researchers introduce Safe-SAIL, a framework that uses sparse autoencoders to interpret safety features in large language models across four domains (pornography, politics, violence, terror). The work reduces interpretation costs by 55% and identifies 1,758 safety-related features with human-readable explanations, advancing mechanistic understanding of AI safety.
AIBearisharXiv – CS AI · Apr 146/10
🧠A research study demonstrates that fine-tuning language models with sycophantic reward signals degrades their calibration—the ability to accurately quantify uncertainty—even as performance metrics improve. While the effect lacks statistical significance in this experiment, the findings reveal that reward-optimized models retain structured miscalibration even after post-hoc corrections, establishing a methodology for evaluating hidden degradation in fine-tuned systems.
AINeutralarXiv – CS AI · Apr 146/10
🧠Researchers introduce SciTune, a framework for fine-tuning large language models with human-curated scientific multimodal instructions from academic publications. The resulting LLaMA-SciTune model demonstrates superior performance on scientific benchmarks compared to state-of-the-art alternatives, with results suggesting that high-quality human-generated data outweighs the volume advantage of synthetic training data for specialized scientific tasks.
AINeutralarXiv – CS AI · Apr 146/10
🧠Researchers present a unified framework for understanding how different methods control large language models—including fine-tuning, LoRA, and activation interventions—revealing a fundamental trade-off between steering strength and output quality. The analysis explains this through an activation manifold perspective and introduces SPLIT, a new steering method that improves control while better preserving model coherence.
AINeutralarXiv – CS AI · Apr 136/10
🧠Researchers introduce PerMix-RLVR, a training method that enables large language models to maintain persona flexibility while preserving task robustness. The approach addresses a fundamental trade-off in reinforcement learning with verifiable rewards, where models become less responsive to persona prompts but gain improved performance on objective tasks.
AINeutralarXiv – CS AI · Apr 106/10
🧠Researchers introduce Sol-RL, a two-stage reinforcement learning framework that combines FP4 quantization for efficient rollout generation with BF16 precision for policy optimization in diffusion models. The approach achieves up to 4.64x training acceleration while maintaining alignment quality, addressing the computational bottleneck of scaling RL-based post-training on large foundational models like FLUX.1.
AIBullisharXiv – CS AI · Mar 276/10
🧠Researchers propose X-OPD, a Cross-Modal On-Policy Distillation framework to improve Speech Large Language Models by aligning them with text-based counterparts. The method uses token-level feedback from teacher models to bridge performance gaps in end-to-end speech systems while preserving inherent capabilities.
AIBearisharXiv – CS AI · Mar 266/10
🧠Research reveals that RLHF-aligned language models suffer from 'alignment tax' - producing homogenized responses that severely impair uncertainty estimation methods. The study found 40-79% of questions on TruthfulQA generate nearly identical responses, with alignment processes like DPO being the primary cause of this response homogenization.
AINeutralarXiv – CS AI · Mar 166/10
🧠Researchers propose Global Evolutionary Refined Steering (GER-steer), a new training-free framework for controlling Large Language Models without fine-tuning costs. The method addresses issues with existing activation engineering approaches by using geometric stability to improve steering vector accuracy and reduce noise.
AINeutralarXiv – CS AI · Mar 66/10
🧠Researchers introduce SalamaBench, the first comprehensive safety benchmark for Arabic Language Models, evaluating 5 state-of-the-art models across 8,170 prompts in 12 safety categories. The study reveals significant safety vulnerabilities in current Arabic AI models, with substantial variation in safety alignment across different harm domains.
AINeutralarXiv – CS AI · Mar 36/1012
🧠RubricBench is a new benchmark with 1,147 pairwise comparisons designed to evaluate rubric-based assessment methods for Large Language Models. Research reveals a significant gap between human-annotated and AI-generated rubrics, showing that current state-of-the-art models struggle to autonomously create valid evaluation criteria.
AINeutralarXiv – CS AI · Mar 36/103
🧠Researchers introduced OVERTONBENCH, a framework for measuring viewpoint diversity in large language models through the OVERTONSCORE metric. In a study of 8 LLMs with 1,208 participants, models scored 0.35-0.41 out of 1.0, with DeepSeek V3 performing best, showing significant room for improvement in pluralistic representation.
