AINeutralarXiv – CS AI · May 16/10
🧠Researchers introduce VISE, the first benchmark for evaluating sycophancy in video large language models (Video-LLMs), where models incorrectly agree with user inputs that contradict visual evidence. The study proposes two training-free mitigation strategies: enhanced visual grounding through keyframe selection and inference-time neural representation steering, addressing a critical reliability gap in multimodal AI systems.
AIBullisharXiv – CS AI · Apr 206/10
🧠Researchers introduce LACE, a framework enabling large language models to reason through multiple parallel paths that interact and correct each other during inference, rather than operating independently. Using synthetic training data to teach cross-thread communication, LACE achieves over 7 percentage points improvement in reasoning accuracy compared to standard parallel search methods.
AINeutralarXiv – CS AI · Apr 206/10
🧠Researchers propose AdaRankLLM, an adaptive retrieval-augmented generation framework that dynamically filters irrelevant passages to reduce computational overhead while maintaining output quality. The study challenges whether adaptive retrieval remains necessary as language models grow more robust, finding that its value differs significantly between weaker and stronger models.
AINeutralarXiv – CS AI · Apr 206/10
🧠Researchers introduce DepCap, a training-free framework that optimizes diffusion language model (DLM) inference through adaptive block-wise parallel decoding. The method achieves up to 5.63× speedup by using cross-step signals to determine block boundaries and identifying conflict-free token subsets for safe parallel execution, maintaining quality while significantly accelerating inference.
AIBullisharXiv – CS AI · Apr 156/10
🧠Researchers propose RPRA (Reason-Predict-Reason-Answer/Act), a framework enabling smaller language models to predict how a larger LLM judge would evaluate their outputs before responding. By routing simple queries to smaller models and complex ones to larger models, the approach reduces computational costs while maintaining output quality, with fine-tuned smaller models achieving up to 55% accuracy improvements.
AINeutralarXiv – CS AI · Apr 146/10
🧠Researchers identify a critical failure mode in non-autoregressive diffusion language models caused by proximity bias, where the denoising process concentrates on adjacent tokens, creating spatial error propagation. They propose a minimal-intervention approach using a lightweight planner and temperature annealing to guide early token selection, achieving substantial improvements on reasoning and planning tasks.
AINeutralarXiv – CS AI · Apr 146/10
🧠StyleBench is a new benchmark that evaluates how different reasoning structures (Chain-of-Thought, Tree-of-Thought, etc.) affect LLM performance across various tasks and model sizes. The research reveals that structural complexity only improves accuracy in specific scenarios, with simpler approaches often proving more efficient, and that learning adaptive reasoning strategies is itself a complex problem requiring advanced training methods.
AINeutralarXiv – CS AI · Apr 146/10
🧠Researchers introduce GroupRank, a novel LLM-based passage reranking paradigm that balances efficiency and accuracy by combining pointwise and listwise ranking approaches. The method achieves state-of-the-art performance with 65.2 NDCG@10 on BRIGHT benchmark while delivering 6.4x faster inference than existing approaches.
AINeutralarXiv – CS AI · Apr 136/10
🧠Researchers introduce Dictionary-Aligned Concept Control (DACO), a framework that uses a curated dictionary of 15,000 multimodal concepts and Sparse Autoencoders to improve safety in multimodal large language models by steering their activations at inference time. Testing across multiple models shows DACO significantly enhances safety performance while preserving general-purpose capabilities without requiring model retraining.
AINeutralarXiv – CS AI · Apr 106/10
🧠Researchers demonstrate that large language models exhibit critical control failures in causal reasoning, where they produce sound logical arguments but abandon them under social pressure or authority hints. The study introduces CAUSALT3, a benchmark revealing three reproducible pathologies, and proposes Regulated Causal Anchoring (RCA), an inference-time mitigation technique that validates reasoning consistency without retraining.
AIBullisharXiv – CS AI · Apr 106/10
🧠Researchers propose a Self-Validation Framework to address object hallucination in Large Vision Language Models (LVLMs), where models generate descriptions of non-existent objects in images. The training-free approach validates object existence through language-prior-free verification and achieves 65.6% improvement on benchmark metrics, suggesting a novel path to enhance LVLM reliability without additional training.
AINeutralarXiv – CS AI · Apr 106/10
🧠Researchers present CGD-PD, a test-time decoding method that improves large language models' performance on three-way logical question answering (True/False/Unknown) by enforcing negation consistency and resolving epistemic uncertainty through targeted entailment probes. The approach achieves up to 16% relative accuracy improvements on the FOLIO benchmark while reducing spurious Unknown predictions.
AIBullisharXiv – CS AI · Apr 106/10
🧠Researchers introduce S³ (Stratified Scaling Search), a test-time scaling method for diffusion language models that improves output quality by reallocating compute during the denoising process rather than simple best-of-K sampling. The technique uses a lightweight verifier to evaluate and selectively resample candidate trajectories at each step, demonstrating consistent performance gains across mathematical reasoning and knowledge tasks without requiring model retraining.
AIBullisharXiv – CS AI · Apr 76/10
🧠Researchers developed a new method to reduce hallucinations in Large Vision-Language Models (LVLMs) by identifying a three-phase attention structure in vision processing and selectively suppressing low-attention tokens during the focus phase. The training-free approach significantly reduces object hallucinations while maintaining caption quality with minimal inference latency impact.
AIBullisharXiv – CS AI · Mar 276/10
🧠Researchers developed InstABoost, a new method to improve instruction following in large language models by boosting attention to instruction tokens without retraining. The technique addresses reliability issues where LLMs violate constraints under long contexts or conflicting user inputs, achieving better performance than existing methods across 15 tasks.
AINeutralarXiv – CS AI · Mar 266/10
🧠Research shows that newer LLMs have diminishing effectiveness for early-exit decoding techniques due to improved architectures that reduce layer redundancy. The study finds that dense transformers outperform Mixture-of-Experts models for early-exit, with larger models (20B+ parameters) and base pretrained models showing the highest early-exit potential.
AIBullisharXiv – CS AI · Mar 176/10
🧠Researchers introduce Truncated-Reasoning Self-Distillation (TRSD), a post-training method that enables AI language models to maintain accuracy while using shorter reasoning traces. The technique reduces computational costs by training models to produce correct answers from partial reasoning, achieving significant inference-time efficiency gains without sacrificing performance.
AIBullisharXiv – CS AI · Mar 176/10
🧠Researchers developed training-free model steering techniques to improve reasoning in large audio-language models (LALMs) through chain-of-thought prompting. The approach achieved up to 4.4% accuracy gains and demonstrated cross-modal transfer where text-derived steering vectors can effectively guide speech-based reasoning.
AIBullisharXiv – CS AI · Mar 176/10
🧠Researchers propose 'Two Birds, One Projection,' a new inference-time defense method for Large Vision-Language Models that simultaneously improves both safety and utility performance. The method addresses modality-induced bias by projecting cross-modal features onto the null space of identified bias directions, breaking the traditional safety-utility tradeoff.
AIBullisharXiv – CS AI · Mar 176/10
🧠Researchers introduce AdaAnchor, a new AI reasoning framework that performs silent computation in latent space rather than generating verbose step-by-step reasoning. The system adaptively determines when to stop refining its internal reasoning process, achieving up to 5% better accuracy while reducing token generation by 92-93% and cutting refinement steps by 48-60%.
AIBullisharXiv – CS AI · Mar 166/10
🧠Researchers propose AdaBoN, an adaptive Best-of-N alignment method that improves computational efficiency in language model alignment by allocating inference-time compute based on prompt difficulty. The two-stage algorithm outperforms uniform allocation strategies while using 20% less computational budget.
AIBullisharXiv – CS AI · Mar 166/10
🧠Researchers introduce Krites, an asynchronous caching system for Large Language Models that uses LLM judges to verify cached responses, improving efficiency without changing serving decisions. The system increases the fraction of requests served with curated static answers by up to 3.9 times while maintaining unchanged critical path latency.
AINeutralarXiv – CS AI · Mar 96/10
🧠Researchers have identified a critical failure mode in Vision-Language-Action (VLA) robotic models called 'linguistic blindness,' where robots prioritize visual cues over language instructions when they contradict. They developed ICBench benchmark and proposed IGAR, a train-free solution that recalibrates attention to restore language instruction influence without requiring model retraining.
AIBullisharXiv – CS AI · Mar 96/10
🧠Researchers introduced VLMQ, a post-training quantization framework specifically designed for vision-language models that addresses visual over-representation and modality gaps. The method achieves significant performance improvements, including 16.45% better results on MME-RealWorld under 2-bit quantization compared to existing approaches.
AINeutralarXiv – CS AI · Mar 45/103
🧠Researchers introduce MELODI, a framework for monitoring energy consumption during large language model inference, revealing substantial disparities in energy efficiency across different deployment scenarios. The study creates a comprehensive dataset analyzing how prompt attributes like length and complexity correlate with energy expenditure, highlighting significant opportunities for optimization in LLM deployment.
