Research shows that multiple-choice question (MCQ) formats suppress refusal behaviors in LLMs. Experiments on Llama-3.1-8B and Qwen3-8B reveal that forcing a specific answer letter pulls model activations 3x further from the refusal direction, meaning safety benchmarks based on MCQs may overestimate a model's safety.
Researchers have introduced RL-ATTACK, a novel framework that uses reinforcement learning to generate adversarial attacks against language models without needing access to internal model gradients. By learning from the target model's responses, this method achieves superior query efficiency and successfully bypasses safety filters by forcing unintended outputs.
Researchers have identified that LLMs suffer from a bias toward providing immediate, often incorrect answers in ambiguous situations instead of asking clarifying questions. They introduced BACE (Belief-Aware Critic with Epistemic tracking), a lightweight method that runs on top of a frozen language model to strategically decide when to request clarification.
Dharma-AI has introduced a 2-billion-parameter model specialized in extracting structured JSON fields from various documents. It supports both open extraction and schema-directed modes, consistently outperforming frontier LLMs on field-level F1 benchmarks while significantly reducing operational costs.
A new method called FlagRot optimizes key-value (KV) cache quantization for long-context models. By selectively rotating key segments based on attention error rather than applying global rotation, it reduces accuracy loss in 2-bit or 4-bit caches.
Researchers have developed a method to calculate the theoretical upper bound for Multi-Draft Speculative Decoding (MDSD) efficiency by solving the dual of the optimal transport problem. The study reveals that current verification algorithms often fall short of this theoretical limit, and that sampling without replacement significantly improves acceptance rates compared to sampling with replacement.
Collinear AI has launched CWE-bench v2, a benchmarking suite featuring 100 tasks covering 91 common vulnerability types. The tool allows developers to programmatically evaluate how effectively their AI agents find and fix security weaknesses in open-source repositories.
Google has deployed a new suite of Cheminformatics Skills for its Antigravity platform, enabling AI agents to manage end-to-end drug discovery pipelines. By integrating the RDKit toolkit with LLM-based orchestration, the system can perform literature research, data analysis, and chemical synthesis planning in approximately one hour.
Research shows that identical persona instructions produce different behavioral outcomes in LLMs when switching from English to Korean. While buyers followed instructions similarly, sellers became significantly more accommodating and polite in Korean, leading to higher agreement rates and lower prices.
Researchers have introduced MEMENTO, a specialized agentic framework built on Qwen3-8B designed to manage long context by compacting history into concise notes. By utilizing fine-tuning on synthetic trajectories followed by reinforcement learning, it achieves high accuracy on book-level tasks while strictly limiting context usage to 20K tokens, even for inputs as large as 2.5 million tokens.
Recent research shows that many attention heads in large language models are redundant and specifically contribute to verbatim memorisation of training data. By selectively pruning these heads at inference time, researchers achieved a 91.5% reduction in memorisation events while simultaneously improving performance on standard benchmarks.
Recent research reveals that language models fail to adapt to evolving user preferences, instead defaulting to 'exploit-first' behavior. While models can identify the correct strategy in multiple-choice scenarios, they show a 35-46% drop in calibration during open-ended generation, persistently over-relying on stale historical data.
NVIDIA has released a new command-line interface for the TensorRT Edge-LLM runtime, allowing developers to quickly prototype and test AI models on the DGX Spark 64GB workstation. The tool chain facilitates local model serving and supports DSpark speculative decoding to boost generation performance on edge-constrained hardware.
Amazon has enabled intent-driven UI automation by pairing Nova Act with the Bedrock AgentCore Browser Tool. This setup allows engineering teams to define test workflows through natural language instructions instead of maintaining rigid, selector-heavy codebases.
IBM has released Granite Switch 4.1.1, a patch update for its modular language model architecture. The update introduces an improved multilingual speech-to-text model and a refined Decision adapter, all achieved without retraining the base model.
Google released a new Agents CLI evaluation namespace designed to help developers perform 'hill-climbing'—an iterative process of improving agent performance through testing, grading, and analysis. The tool enables regression tracking by comparing agent traces across model swaps or prompt changes, surfacing issues like tool-use regressions that standard unit tests miss.
Researchers have proposed LOTUS, a framework for hallucination detection that uses multilinear low-rank decomposition to extract signals from hidden LLM layers. By analyzing peak activations and cross-layer patterns, the method identifies incorrect responses more effectively than existing pooling-based techniques.
Google Project Zero analyzed how LLM-based systems (BigSleep) compare to traditional generative fuzzers (Fuzzilli) in finding V8 JavaScript engine vulnerabilities. Over six months, both methods independently discovered unique bugs, with an estimated overlap of approximately 25-60%.
LiquidAI has launched LFM3, a new generation of small foundation models built on a shared 2B-parameter backbone. The series includes text, vision, audio, and omni models designed for high-efficiency on-device deployment.
NVIDIA has released 'Nemotron-Personas' datasets for France and Belgium, designed to improve LLM accuracy regarding regional procedures, demographics, and cultural nuances. The update provides a complete post-training workflow—including data generation, LoRA fine-tuning via NVIDIA NeMo, and evaluation—to help developers transition from generic models to domain-specific agents.
Researchers have introduced the Boa Compressor, an online tool utilizing a Mamba state-space model to compress long-form context into shorter representations in linear time. By distilling knowledge from full-context LLM interactions, this method allows for efficient per-query inference without the massive latency overhead of transformer-based compression.
Researchers introduced Recursive Semantic Divergence (RSD), an unsupervised metric that measures and penalizes inconsistency in LLM agent trajectories. By applying RSD as a reward signal during fine-tuning, models exhibit significantly more stable behavior and higher success rates across tool-use benchmarks.
Recent research on Qwen3.5-27B demonstrates that while activation interventions can force a model to produce a specific target output, they often fail to instill correct underlying reasoning. Researchers found that gains in target selection collapse when facts are recombined, suggesting that forced outputs do not reliably represent valid tool-use condition logic.
Audible has developed a controllable text-to-speech pipeline that uses LLMs to profile characters and attribute dialogue with up to 97% accuracy. By providing 'stage directions' as natural-language captions, the system ensures distinct, consistent character voices over long-form content.