AIBullisharXiv – CS AI · Jun 27/10
🧠Researchers introduce TAPS, a target-aware prefix selection method that improves speculative decoding by optimizing how draft trees are verified in diffusion models. The technique achieves up to 7.9x speedup over standard autoregressive decoding and outperforms competing methods by 1.36-1.74x, addressing a fundamental inefficiency where existing approaches verify unreachable token sequences.
AIBullisharXiv – CS AI · Jun 27/10
🧠Researchers demonstrate that 2-bit quantization of large reasoning models causes instability leading to longer inference traces rather than speedup, but introduce lightweight recovery techniques (FP16 planning and loop rescue) that restore accuracy from 17-65% to 74-87% while maintaining computational efficiency.
AIBullisharXiv – CS AI · Jun 27/10
🧠Researchers introduce DeLask, a novel decoding framework that reduces hallucinations in Large Language Models by dynamically skipping decoder layers prone to generating false information. The method uses gradient-based analysis to identify problematic layers and partially aggregates their hidden states, demonstrating consistent improvements across diverse LLMs without requiring model retraining.
AIBullisharXiv – CS AI · Jun 27/10
🧠Xiaomi researchers have developed MiCU, a domain-specific large language model optimized for smart home command understanding that handles ambiguous user requests better than traditional systems. The model employs curriculum learning, reinforcement learning, and token compression techniques, achieving 20% average accuracy gains and reducing user correction rates by 1.57% in production deployment across 1.7 million daily active users in the Xiaomi Home app.
AIBullisharXiv – CS AI · Jun 17/10
🧠Researchers propose MedCoG, a meta-cognitive agent that improves Large Language Model efficiency in medical reasoning by dynamically regulating knowledge utilization based on self-assessed task complexity and familiarity. The approach achieves 6.2x inference density improvement while reducing computational costs and improving accuracy on medical benchmarks.
AIBullisharXiv – CS AI · May 297/10
🧠Researchers introduce HARP, a learnable adaptive rotation processor that improves extreme low-bit quantization for large language models by replacing fixed Hadamard transforms with optimizable structured orthogonal processors. The technique maintains full-precision equivalence while achieving better perplexity and accuracy across 2-4 bit quantization settings on models up to 70B parameters, with deployment speeds competitive with standard approaches.
🏢 Perplexity
AIBullisharXiv – CS AI · May 297/10
🧠Researchers introduce CORE-RAG, a novel framework that compresses context in Retrieval-Augmented Generation systems using performance-driven learning rather than predefined heuristics. The approach achieves a 97% compression ratio while improving accuracy by 3.3 points on exact match scores, addressing a critical bottleneck in LLM efficiency.
AIBullisharXiv – CS AI · May 297/10
🧠Researchers propose BRACS, a training-free framework that reduces hallucinations in vision-language models by monitoring visual grounding during text generation and applying adaptive corrections only when needed. The method achieves significant improvements on hallucination benchmarks while maintaining computational efficiency comparable to baseline decoding speeds.
AIBullisharXiv – CS AI · May 297/10
🧠Researchers have developed a method to improve how large language models verify factual claims by framing fact-checking as a true/false reading comprehension task with explicit test-taking strategies. The approach reduces token usage by over 80% while maintaining competitive performance, and enables smaller language models to perform similarly to larger ones through fine-tuning and self-revision mechanisms.
AIBullisharXiv – CS AI · May 297/10
🧠Researchers propose DenseSteer, a training-free framework that improves mathematical reasoning in small language models (≤3B parameters) by steering internal representations toward denser reasoning patterns. The method demonstrates that smaller models can match larger ones' performance by executing fewer, more information-rich reasoning steps rather than verbose chain-of-thought processes.
AIBullisharXiv – CS AI · May 297/10
🧠Researchers introduce COLAGUARD, a new safety guardrail system for large language models that embeds multi-step reasoning into latent space, achieving comparable safety performance to explicit reasoning models while delivering 12.9X faster inference and 22.4X reduction in token usage. The approach addresses a critical bottleneck in deploying AI safety systems at scale by eliminating the computational overhead of traditional reasoning-based content moderation.
🧠 Llama
AIBullisharXiv – CS AI · May 297/10
🧠Researchers introduce LoRe, a training-free optimization method that dynamically routes computational resources to high-priority interactions in iterative graph solvers, achieving 8× speedup and 12× memory reduction on combinatorial optimization problems while maintaining solution quality.
AIBullisharXiv – CS AI · May 287/10
🧠Researchers introduce EAGer, a training-free method that optimizes inference-time computation for reasoning language models by dynamically allocating compute budgets based on token-level entropy. The approach reduces computational waste while improving performance, achieving up to 37% gains in Pass@k metrics with 59% fewer tokens in supervised settings.
AIBullisharXiv – CS AI · May 287/10
🧠Researchers propose Group-Query Latent Attention (GQLA), an advancement of DeepSeek's Multi-head Latent Attention that enables hardware-adaptive decoding through two algebraically equivalent inference paths without requiring model retraining. The innovation allows a single trained model to optimize performance across different hardware platforms—H100 GPUs and export-restricted H20 chips—while maintaining computational efficiency and supporting distributed tensor parallelism.
AIBullisharXiv – CS AI · May 287/10
🧠Researchers present a guided stochastic exploration framework that enhances inference in recursive neural network architectures by treating reasoning as approximate inference over latent trajectories. The method uses stochastic perturbations and model-based reweighting to improve performance on structured reasoning tasks, achieving 98% accuracy on Sudoku-Extreme (up from 85.9%) while providing three label-free diagnostics to assess reliability without retraining.
AIBullisharXiv – CS AI · May 287/10
🧠Researchers propose Locality-Aware Redundancy Pruning (LoRP), a training-free method for compressing large language models by removing redundant layers based on representational similarity patterns. The framework uses a Representation Locality Score to identify and prune depth-wise redundancy more effectively than existing approaches, improving both perplexity and downstream task performance across multiple LLM architectures.
🏢 Perplexity
AIBullisharXiv – CS AI · May 287/10
🧠Researchers propose a sleep-like mechanism for transformer language models that periodically consolidates context into persistent fast weights, reducing the computational burden of long sequences. The method shifts heavy computation offline while maintaining fast inference speeds, showing significant improvements on reasoning tasks that standard transformers struggle with.
AIBullisharXiv – CS AI · May 287/10
🧠Researchers present a framework for converting Mixture-of-Experts (MoE) language models into standard dense architectures through expert selection, grouping, and knowledge distillation. The method achieves superior performance compared to traditional dense-to-dense pruning while enabling deployment on memory-constrained systems.
AIBullisharXiv – CS AI · May 287/10
🧠Researchers introduce CIVIC, a framework that optimizes Vision-Language Models by maintaining compact visual token sequences throughout the entire inference pipeline, reducing KV-cache memory to one-third while achieving measurable hardware acceleration without accuracy loss.
AIBullisharXiv – CS AI · May 287/10
🧠Researchers propose a new runtime layer architecture for serving multi-agent LLM systems, positioned between application frameworks and inference engines. The approach enables unified policy management for cross-cutting concerns like caching and fairness, with CacheSage demonstrating 13-37% improvements in cache hit rates and 12-29% reductions in time-to-first-token latency.
AIBullisharXiv – CS AI · May 287/10
🧠GroundedCache proposes a safety-first framework for reusing cached answers in retrieval-augmented generation systems by validating four conditions before serving cached responses. The system achieves near-zero unsafe-served rates (0-1.5%) across benchmarks while maintaining minimal latency overhead, addressing critical vulnerabilities in current caching approaches that can serve incorrect answers.
AIBullisharXiv – CS AI · May 277/10
🧠Researchers demonstrate that stochasticity in discrete diffusion models provides an error-correcting mechanism that improves the speed-quality tradeoff in generative AI. They propose Discrete Churn and Restart Sampling (DCRS), which achieves up to 10x faster sampling on images while maintaining quality by strategically injecting controlled randomness into the inference process.
AINeutralarXiv – CS AI · May 277/10
🧠Researchers introduce ICCU, an in-context continual unlearning framework that removes specific data influence from language models without modifying parameters. The method uses pattern-induced refusal rules applied at inference time, addressing the inefficiency of sequential unlearning requests in production deployments.
AIBullisharXiv – CS AI · May 277/10
🧠Researchers introduce JetViT, a hybrid Vision Transformer architecture that maintains accuracy of state-of-the-art models while delivering up to 1.79x faster throughput and 44.81% lower latency on high-resolution images. The innovation uses post-training attention search to convert full-attention models into efficient hybrid variants by strategically replacing redundant attention blocks.
🏢 Nvidia
AIBullisharXiv – CS AI · May 277/10
🧠Researchers introduce Self-Signals Driven Multi-LLM Debate (SID), a method that leverages internal model signals like token logits and attention mechanisms to improve multi-agent LLM reasoning while reducing computational overhead. The approach enables high-confidence models to exit early and compresses redundant debate content, achieving better accuracy with lower token consumption than existing multi-LLM debate techniques.
