Tetra introduces a new Leech‑lattice based codebook that reduces the memory footprint of quantized LLMs to about 2.15 bits per weight, enabling efficient 2‑bit quantization without a massive lookup table. The method employs a 64‑state Golay trellis and a shared 16 KiB table, decoding each 24‑weight block with only six table loads and two small lookups. When applied to Qwen3 models (4B, 8B, 14B), Tetra achieves 2.70–2.73 bits per parameter, scoring 63–75 on MMLU and generating 57–114 tokens per second, while maintaining close performance to 4‑bit AWQ and outperforming llama.cpp’s IQ2_XXS on 4B.
AutoBCI is an agentic framework that uses a Designer Agent and a Forecaster Agent to discover and select EEG decoding architectures across diverse tasks such as emotion recognition, motor imagery, and sleep staging. The Designer Agent performs Pool‑Guided Architecture Discovery (PGAD) to generate and refine models, while the Forecaster Agent uses Performance Estimation from Early Knowledge (PEEK) to predict full‑budget validation performance from early learning curves. In experiments on 14 EEG datasets, AutoBCI with Claude Opus 5.5 achieved a 64.16% average test balanced accuracy, slightly surpassing the best baseline, and PEEK reduced prediction error by 38.1% compared to the best-observed-score baseline.
The paper investigates how hybrid attention mechanisms—combining full softmax attention with recurrent alternatives—affect multilingual language models, especially for long sequences and poorly tokenized languages. Interpretability analysis reveals that cross‑lingual representations form patterns linked to the ordering of recurrent and full‑attention layers, with a notable spike in alignment after the first full‑attention layer. Distillation experiments show that alternative layer orderings consistently outperform the standard arrangement, achieving up to 2.5× faster learning, suggesting that starting with a full‑attention layer may benefit multilingual models.
The paper evaluates whether incorporating the hierarchical structure of Tironian notes can improve automatic recognition of this complex Latin shorthand system. Experiments compare flat classifiers (ResNet18, ConvNeXt, Swin, ViT) with hierarchy‑aware models (HD‑CNN and routing approaches) on handwritten and manuscript samples, with and without few‑shot adaptation. Results show that hierarchical models outperform flat ones when no adaptation is applied, but flat models surpass them after few‑shot adaptation, indicating that hierarchy can aid recognition under non‑adapted conditions.
The article explains how to transform a small open‑source Qwen LLM into a fast, single‑pass text classifier by replacing its language‑modeling head with a JEV model. It provides a step‑by‑step guide to swapping the head, enabling the LLM to perform classification tasks efficiently. The process leverages the flexibility of open‑source models to create a lightweight, high‑performance classifier.
By Anubhab Banerjee
The paper presents an expert-driven method for turning normative principles—specifically Islamic ethical, theological, and jurisprudential traditions—into alignment data for language models. Over a year, seven experts curated 2.8K supervised fine-tuning examples and 5.4K preference pairs in Arabic-English, then evaluated models trained on these datasets. Experiments show that models trained with the curated SFT data outperform a baseline in expert judgments, while adding preference data yields a smaller, non-significant improvement.
Timeline-Bench is a benchmark comprising 56 real video‑editing tasks that require AI agents to transform raw production material into finished videos. Each task includes a brief, source assets, a container, and a set of tests that assess format, content, brief compliance, and quality based on 2,582 blind judgments by 43 video editors. In evaluations, the best agent resolved only 15 of the 56 tasks, and most failures were due to quality tests rather than technical errors.
The article titled "Import AI 474: Platonic mindspace; TPUs in space; Zhipu starts an outer RSI loop" discusses the limits of large language models (LLMs) and explores scenarios where their capabilities are surpassed. It highlights emerging technologies such as TPUs deployed in space and the initiation of an outer RSI loop by Zhipu, suggesting new frontiers in AI development. The piece invites readers to consider where current LLMs fall short and what innovations might bridge those gaps.
By Jack Clark
The paper investigates post‑training quantization of transformer attention blocks by optimizing a joint loss over the Q, K, V projections rather than individual weight matrices. Using this joint attention‑based objective (JAB), the authors achieve significant compression on Mistral‑7B, recovering 77‑90% of the performance gap at 3 bits, but the method fails when MLP layers are included. A role‑aware offset rule that ignores sensitivity estimates outperforms JAB on GPT‑2 and full Mistral‑7B, demonstrating that the matrix a weight belongs to is more critical than sensitivity metrics.
The paper introduces a training‑free visual token pruning strategy for vision‑language models that separates early vision‑guided pruning from later text‑guided reselection. By first pruning tokens with vision‑encoder attention, retaining candidates until the decoder midpoint, and then applying text‑to‑visual attention, the method preserves task‑relevant visual information. Across eight benchmarks and three models, it achieves an average performance recovery of 11.10 and 16.84 percentage points at 80% and 90% pruning, respectively, while maintaining comparable or lower LLM‑prefill latency.
AgentPerfBench is a new benchmarking suite designed to evaluate the inference performance of agentic large language models (LLMs) that handle multi‑turn, tool‑using, and context‑expanding tasks. It builds on real traces from agentic benchmarks such as SWE‑Bench and TerminalBench, and generates synthetic profiles that reflect realistic input/output lengths and turn counts. The suite also provides kernel‑level Nsight Compute traces and a multi‑dimensional roofline model to identify hardware bottlenecks and quantify the gap between traditional chat benchmarks and agentic workloads.
The paper introduces a dynamic bipartite matching framework that uses large language model (LLM) agents and contextual bandits to model decentralized, asynchronous matching processes without requiring full preference rankings. In a simulated Chinese marriage market, LLM agents evaluate local candidates while Logistic-UCB models learn reciprocal acceptance, leading to higher mutual welfare and fewer blocking pairs compared to classical Gale–Shapley. The study validates LLM-generated preferences against empirical data and demonstrates gender-differentiated acceptance patterns, supporting the use of decentralized LLM-based matching for economic simulation and computational social science.
PMOPD introduces a projection-based approach to multi-teacher on-policy distillation, addressing the capability seesaw problem by constructing subspace memories from task-specific parameter displacements and projecting gradients and optimizer updates to avoid cross-task interference. It also includes a lightweight conflict probe for task interaction analysis, a task ordering strategy, and a cycling mechanism to balance subspace estimation and task revisitation. Experiments on Code, Reason, and Math tasks demonstrate that PMOPD improves all evaluated capabilities, raising average scores by 2.54 points on Qwen2.5-7B and 2.09 points on Llama-3.1-8B.
EOPSA (Efficient On-Policy Self-Distilled Safety Alignment) addresses inefficiencies in On-Policy Self-Distillation (OPSD) for safety alignment by focusing training on safety-critical tokens. It introduces Adaptive Rollout Scheduling, which limits generation length based on a Teacher Rescue Rate metric, and Selective Distillation, which filters out safety-neutral tokens to concentrate gradient updates on safety-pivotal transitions. Experiments on models up to 32B parameters show that EOPSA reduces rollout computation by about 50% and backpropagates through only roughly 2% of tokens, outperforming full-token distillation baselines in safety compliance and reasoning retention.
Masked diffusion language models can generate outputs in any order, but recent findings suggest this flexibility may reduce diversity by delaying uncertain tokens. The study identifies low‑confidence remasking (LCR) as the main culprit, showing that its filtering of lower‑probability tokens suppresses diversity exponentially. Replacing LCR with top‑probability position selection (TPP) restores diversity, and adding Entropy‑Guided Initialization (EGI) further enhances rollout diversity and solution coverage, demonstrating the benefits of arbitrary‑order generation for diverse outputs.
KiT is a K‑line Diffusion Transformer foundation model designed for financial time‑series forecasting. It reframes future prediction as conditional path generation via flow matching, producing ensembles of plausible OHLCV trajectories from a historical context window. Trained on billions of candlestick bars across multiple markets and timescales, KiT achieves superior RankIC scores compared to task‑specific forecasters and general time‑series models.
M3OS is a multi‑agent large‑language‑model system that separates molecular‑design reasoning from optimization‑state management using a Monte Carlo graph search. The system maintains a persistent graph of evaluated candidates, transformations, and evidence, while LLM agents use role‑specific contexts to generate and edit molecules with tool‑driven and knowledge‑guided approaches. Across three benchmarks, M3OS outperforms baselines, demonstrating the benefit of persistent search state, specialized agents, and controlled execution for multi‑constraint molecular optimization.
The paper introduces ATLAS, a method that uses retained-domain activation atlases to guide task-specific fine‑tuning of language models. By providing local reference centers and directional filters, ATLAS learns a low‑rank residual that adapts to new tasks while preserving existing behavior. Experiments on Qwen3‑8B and other backbones show lower retained‑output KL divergence, fewer rewritten answers, and more stable responses compared to seven baselines.
RGDT-Bench is a new benchmark that evaluates large language models on Rule‑Governed Decision Tasks, where models must apply external rules to facts, justify decisions, and provide checkable justifications. The benchmark offers 202.1K condition‑level supervision slots across four task tracks and eight task‑probe combinations, and it labels warrant completeness through label‑blind extraction and deterministic checks. Evaluation shows that among correct responses, 40.2% of warrants are incomplete, and existing evaluators struggle to detect this, prompting the authors to train a reward model that improves AUROC to 69.24% and outperforms outcome‑supervised baselines.
The paper introduces Q-learning Penalized Transformer (QPT), a training–inference consistent framework for safe offline reinforcement learning. QPT trains a Transformer policy that generates actions conditioned on trajectory context and target return/cost while incorporating a Q-shaped penalty to balance safety, reward maximization, and behavior regularization. The method consistently outperforms strong baselines on 38 DSRL benchmark tasks and adapts robustly to varying constraint thresholds.