AIBullisharXiv – CS AI · Jun 257/10
🧠Researchers demonstrate that multi-agent document assessment for retrieval-augmented generation (RAG) systems can be significantly optimized through model-adaptive routing rather than expensive scoring mechanisms. The study reveals that weaker models benefit primarily from document isolation rather than quality assessment, while MADARA, a proposed adaptive architecture, generalizes across different model families with zero-shot capability, reducing computational overhead.
AIBullisharXiv – CS AI · Jun 237/10
🧠Researchers introduce Self-Aware Scheduling (SAS), a method that learns optimal token unmasking orders in masked diffusion language models through policy optimization. The approach significantly improves generation quality on reasoning tasks, achieving 91.8% accuracy on Sudoku (up from 82%) and boosting mathematical reasoning performance by 12 percentage points on GSM8K.
AIBullisharXiv – CS AI · Jun 237/10
🧠Researchers introduce CORE, a lightweight prompt compression method that optimizes large language models for edge devices without requiring auxiliary smaller models. The approach achieves 30% accuracy improvements while reducing memory usage by 50% and cutting energy consumption by 95% on smartphones compared to existing methods.
🏢 Nvidia
AIBullisharXiv – CS AI · Jun 237/10
🧠Researchers prove theoretically that reinforcement learning with verifiable rewards (RLVR) enables language models to learn efficient backtracking strategies superior to supervised fine-tuning (SFT), achieving exponential computational advantages during inference. The study models chain-of-thought reasoning as graph pathfinding and demonstrates that RLVR trains models to identify difficult decision points, allowing better allocation of compute resources.
AIBullisharXiv – CS AI · Jun 237/10
🧠HyperQuant is a new post-training quantization pipeline that compresses large language and diffusion models to 3-5 bits per weight while maintaining near-lossless quality, outperforming existing methods like HIGGS and TurboQuant. The technique combines Hadamard transforms, optimal lattice quantization, and entropy coding to achieve 3.9x compression on model weights and 3.79x on KV cache, enabling more efficient deployment of large AI models.
AIBullisharXiv – CS AI · Jun 237/10
🧠Researchers introduce Explore-Execute Chain (E²C), a structured reasoning framework that separates LLM planning from execution into distinct computational phases. The approach achieves 53.3% accuracy on AIME 2024 benchmarks with significantly fewer tokens than existing methods, while enabling efficient domain adaptation through exploration-focused fine-tuning.
AIBullisharXiv – CS AI · Jun 117/10
🧠FlowBank presents a novel framework for optimizing LLM-based multi-agent systems by building a portfolio of complementary workflows rather than searching for a single universal solution or regenerating workflows per query. The approach balances computational efficiency with performance, achieving 4-14% improvements over existing methods while reducing inference costs.
AIBullisharXiv – CS AI · Jun 117/10
🧠Researchers introduce HORMA, a hierarchical memory system for LLM agents that organizes experience into structured hierarchies with linked summaries and raw trajectories. The system achieves 22% token efficiency on long tasks while maintaining performance, addressing critical limitations in how language model agents manage working memory for multi-step reasoning.
AINeutralarXiv – CS AI · Jun 107/10
🧠Researchers discovered that key-value cache quantization—a technique used to reduce LLM inference memory—silently degrades AI safety alignment without affecting standard performance metrics like perplexity. The study identifies the root cause as geometric vulnerability of safety features in low-dimensional activation subspaces and proposes Per-Channel Reduction (PCR), a diagnostic tool that achieves up to 97% alignment recovery without retraining.
🏢 Nvidia🏢 Perplexity
AIBullisharXiv – CS AI · Jun 107/10
🧠Researchers introduce PiSO (Piecewise Scale Optimization), an algorithm that optimizes quantization scaling factors for compressing large language models more effectively than existing heuristic methods. By using calibration data to compute optimal channel-wise scales, PiSO demonstrates consistent improvements in model perplexity and downstream accuracy across Llama and Qwen models, with gains becoming more pronounced at lower bit-widths.
🏢 Perplexity🧠 Llama
AIBullisharXiv – CS AI · Jun 107/10
🧠Researchers have mapped how Audio-Visual Large Language Models (AVLLMs) process and integrate audio and visual information internally, revealing distinct information flow patterns depending on input configuration. The study demonstrates that multimodal tokens can be pruned after information transfer with minimal performance impact, enabling more efficient inference across different model scales.
AIBullisharXiv – CS AI · Jun 97/10
🧠Researchers demonstrate that smaller language models (270M-8B parameters) can match or nearly match the performance of larger models for merchant information extraction in financial transactions through strategic fine-tuning techniques. The study identifies Qwen 3.5 4B as achieving 96.60% F1 score with half the parameters of the baseline LLaMA 3.1-8B model, offering significant cost and latency improvements for production deployment.
AIBullisharXiv – CS AI · Jun 97/10
🧠Researchers introduce Collaboration Policy Tree (Co-pi-tree), a method that distills large language model reasoning into interpretable, executable policy trees for human-AI collaboration. The approach achieves 35% performance improvement while reducing LLM queries by 78% and latency by 97%, addressing key limitations of black-box reinforcement learning and costly real-time LLM querying.
AIBullisharXiv – CS AI · Jun 97/10
🧠Researchers introduce EntropyInfer, a training-free framework that optimizes long-context LLM inference by dynamically allocating computational resources based on attention entropy patterns. The method achieves up to 2.39× speedup on models like Llama and Qwen beyond 100k tokens while maintaining output quality, addressing limitations in existing sparse attention and KV cache compression techniques.
🧠 Llama
AIBullisharXiv – CS AI · Jun 97/10
🧠ScaleSweep introduces an optimized block scale initialization method for NVFP4 quantization of large language models, improving upon traditional AbsMax approaches. The technique theoretically bounds the search space and empirically achieves 93% performance retention under aggressive 4-bit quantization, advancing hardware-efficient AI inference.
🧠 Llama
AIBullisharXiv – CS AI · Jun 97/10
🧠Researchers introduce FlashMemory-DeepSeek-V4, a novel inference system using Lookahead Sparse Attention to reduce GPU memory requirements for long-context LLM serving by 86.5% while maintaining accuracy. The approach uses a neural memory indexer to selectively preserve only critical KV cache chunks, enabling efficient processing of ultra-long contexts up to 500K tokens.
AIBullisharXiv – CS AI · Jun 87/10
🧠Researchers introduce SPpruner, a new vision-language model optimization technique that reduces computational costs by intelligently filtering visual tokens while maintaining accuracy. The method achieves up to 2.53x speedup with minimal performance loss by prioritizing semantically relevant subjects and their contextual relationships, addressing a major bottleneck in VLM inference.
AIBullisharXiv – CS AI · Jun 57/10
🧠Researchers introduce ReLAT, a test-time training method that improves latent reasoning in large language models by reconstructing the original query from intermediate latent states, ensuring task-relevant information is preserved. The approach demonstrates significant performance gains across mathematical reasoning, QA, and code generation tasks, with Qwen3-8B achieving a 16.6-point improvement on AIME 2024.
AIBullisharXiv – CS AI · Jun 47/10
🧠Researchers introduce YAQA, a new quantization algorithm that improves model compression by directly optimizing end-to-end error rather than layer-by-layer error. The method achieves 30% error reduction compared to existing approaches like GPTQ and even outperforms quantization-aware training, with theoretical guarantees backing its performance.
AIBullisharXiv – CS AI · Jun 37/10
🧠Researchers introduce Reasoning Primitive Induction, a method that extracts reusable reasoning patterns from ReAct-style LLM agent traces and converts them into a compact library of pseudo-tools. The induced libraries consistently outperform the original agents by 22-44 percentage points across multiple reasoning tasks, suggesting a systematic path to improve LLM reasoning through learned decomposition.
AIBullisharXiv – CS AI · Jun 27/10
🧠Researchers introduce SubFit, a post-training compression method for Large Language Models that operates at the submodule level rather than full-layer granularity, achieving superior perplexity-accuracy trade-offs. The approach selects non-contiguous Attention and FeedForward submodules with individual fitted residual bypasses, delivering 84.6% downstream accuracy retention at 25% sparsity compared to 81.6% for existing methods.
🏢 Perplexity
AIBullisharXiv – CS AI · Jun 27/10
🧠LayerRoute is a lightweight adapter that enables language models to dynamically skip transformer blocks based on input type, achieving 12.91% computational efficiency gains with minimal training overhead. By combining per-layer routers with LoRA fine-tuning, the system learns to skip 15.25% of computations for tool calls while maintaining full capacity for complex reasoning tasks, demonstrating significant potential for optimizing agentic AI systems.
🏢 Perplexity
AIBullisharXiv – CS AI · Jun 27/10
🧠BudgetDraft is a new training method for sparse-KV speculative decoding that enables faster language model inference under memory constraints. By training drafters to handle multiple KV cache budgets simultaneously, the technique achieves up to 6.55x speedup on mid-to-long context inference while maintaining acceptance rates and reducing GPU memory usage.
AIBullisharXiv – CS AI · Jun 27/10
🧠Researchers introduce WaveFilter, a training-free framework that uses wavelet transforms to optimize Key-Value cache filtering in Diffusion Large Language Models, addressing computational bottlenecks in long-context processing. The technique enables sparse KV caching to maintain generation quality while reducing inference latency, offering plug-and-play compatibility with existing LLM architectures.
AIBullisharXiv – CS AI · Jun 27/10
🧠Researchers propose LU-KV, a novel framework for optimizing KV cache eviction in large language models by formulating budget allocation as a combinatorial optimization problem. The approach reduces KV cache size by 80% while maintaining performance, significantly lowering inference latency and GPU memory requirements.
