FastCI is a new framework designed to make continuous integration (CI) for large language model (LLM) training frameworks more efficient. It uses runtime evidence to select and prune tests that run changed code in equivalent contexts, prioritizes high‑risk tests, and optimizes test workloads beyond each test’s intended validation scope. In evaluations on ByteDance’s LLM training framework, FastCI cut CI latency by 77.5 % and GPU usage by 63.9 %, while slightly improving modified code coverage retention.
By Tianshuo Qiao, Naiqian Zheng, Xiaopeng Liu, Shuguang Wang, Diandian Gu, Xuanzhe Liu, Xin Jin
The paper audits whether synthetic distractors in RLVR corpora act as shortcuts for learning policies. A classifier using only surface statistics barely outperforms chance, and manual inspection reveals that code distractors are almost identical to correct answers. Experiments with a paraphrase‑matched control show no exploitation advantage for the unmodified data, indicating that the detectable artifact was not used by the policy.
By Esther Xin
The study evaluates whether aggregated brand recommendation profiles can identify the language model that generated them. Using 6,475 responses from five deployed endpoints, a character‑n‑gram classifier accurately attributes single responses to the correct system (97.84% accuracy). However, when responses are aggregated into domain‑condition units, the classifier’s performance drops to 66.53%, and a forest model misclassifies all gift‑domain units, indicating that aggregated brand behaviour does not reliably reveal the underlying system.
By Dmitrij \.Zatuchin
The paper investigates how fine‑tuning large language models with a small number of harmful examples can erode their refusal behavior, and explores whether localizing safety‑related behavior to specific layers or directions can provide robust defenses. Experiments across six checkpoints from four model families show that harmful and benign prompts remain linearly separable after attack, and that patching clean hidden states or freezing layers up to a transition depth can restore refusal. However, attackers can bypass these defenses by spreading updates or targeting singular directions, indicating that adaptive fine‑tuning can defeat localized repairs and highlighting the need for multiple defensive checks.
By Jungseob Lee, Dongyub Jude Lee, Sugyeong Eo, Seongtae Hong, Seungyoon Lee, Heuiseok Lim
The paper introduces NEEDLE, a training‑free technique for removing backdoors from large language models. After a trigger is identified, NEEDLE estimates a backdoor direction and a refusal subspace using activation vectors, then applies sequential weight orthogonalisation to suppress the backdoor while preserving refusal‑related representations. The method requires no clean reference model or original poisoned data and achieves the lowest attack success rate and minimal impact on model performance across multiple model families and attack types.
By Minoo Kim, Vasileios Lampos, George Drayson
AIR-LLM is an edge inference architecture that broadcasts large language model (LLM) weights over radio, allowing edge devices to perform matrix-vector multiplications directly in the RF domain without storing or loading the weights. The system uses MIMO spatial multiplexing and an energy‑efficient precoder‑postcoder pair to reduce airtime and calibrate the wireless channel, enabling a single broadcast to serve unlimited users. Experiments on real urban channel models show that AIR-LLM achieves only a 4.0% perplexity loss on LLaMA‑3.1‑8B while saving energy by up to 157.7× compared to FP16 and reducing airtime by over 100× for 20 users.
By Zhihui Gao, Tingjun Chen, Dirk Englund
Leto is a fault‑tolerant training system for large language models that enables fast in‑place recovery on surviving hardware after hardware‑operable failures. It retains the working model state and reusable process state, while pre‑initializing remaining state in a shadow trainer, using two‑tier erasure protection and chunk‑level transactional updates to maintain consistency. Experiments on NVIDIA A100 clusters show Leto recovers 3.6–6.5× faster than checkpointing baselines and boosts productive training time by up to 13.7 percentage points, with simulations indicating over 95% productivity on a 131,072‑GPU cluster.
By Geon-Woo Kim, Joon Ha Kim, Daehyeok Kim
STEER is a sampling method for relational foundation models that reduces inference cost by focusing on the most relevant tables for a prediction task. It uses a large language model to rank foreign‑key edges in the database schema into relevance tiers, then assigns traversal probabilities based on these tiers. Evaluated on three state‑of‑the‑art RFMs, STEER cuts inference context size by roughly 40% on average while preserving or improving accuracy.
By Abdalla Mohamed, Ashraf Aboulnaga
The paper investigates why Group Relative Policy Optimization (GRPO) benefits from per‑prompt normalization by examining the local curvature of the sequence‑level policy gradient. It shows that standard deviation normalization acts as an adaptive gradient, yielding a provably faster convergence rate than unnormalized REINFORCE under mild conditions, with the improvement tied to the average within‑prompt reward standard deviation. The authors also propose IS‑GRPO, an importance‑sampling variant that maintains alignment with the full gradient and offers a tighter convergence guarantee, and empirically validate these theoretical insights on GSM8K and MATH datasets at 1.5B and 7B model scales.
By Cheng Ge, Caitlyn Heqi Yin, Hao Liang, Jiawei Zhang
GLoC-EHR is a multimodal language model that processes electronic health records by combining a fixed-size global memory of the entire patient trajectory with a local memory of selected events. It generates hospital-course summaries and masked concept descriptions, then is fine‑tuned to cite evidence before answering clinical questions, using group relative policy optimization to reward correct, evidence‑supported responses. On MIMIC‑IV outcome tasks, GLoC‑EHR achieves the highest macro AUROC among compared models when answering directly, and maintains strong performance with evidence‑cited reasoning while adding distinct supported findings from the local memory.
By Chaiho Shin, Kwangsoo Kim
arXiv:2610. 01172v1 Announce Type: new Abstract: We study learning rate (LR) scaling for hybrid architectures combining Transformer and State-Space Model (SSM) blocks, a class adopted by several recent production language models.
By Jimin Seo, Gyubok Lee, Yeonsik Jo, Kiwoong Yoo, Yeongoon Kim, Minhae Oh, Jin Woo Koo, Suhwan Kim, Nakyung Lee, Minsik Seol, Idris Nechnech, Jaehyeon Kim, Giho Lee, Jungwoo Lee
The paper introduces TACO, a new optimizer for fine‑tuning large language models that drastically reduces optimizer state memory while preserving first‑order gradients. TACO selects the sign of the largest magnitude entry in each column of weight matrices, achieving a 174× reduction in persistent optimizer memory compared to AdamW8bit and a 2.9× decrease in peak training memory on OPT‑13B. This allows full‑parameter fine‑tuning of 30–32B‑parameter models on a single 80 GB GPU across multiple model families and tasks, with comparable accuracy and runtime to existing methods.
By Jichao Jiang (University of Central Florida), Cristian McGee (University of Central Florida), El Houcine Bergou (Mohammed VI Polytechnic University), Hanqin Cai (University of Central Florida), Aritra Dutta (University of Central Florida)
The paper introduces an agentic forecasting environment built on 2,100+ resolved Polymarket questions, where a language model (Qwen3.5-35B-A3B) learns to gather evidence during rollout via web search, page reading, and financial time series, all filtered to avoid post‑cutoff leaks. Training with single‑epoch GRPO and a Brier‑score reward improves calibration by 30‑40% and reduces search attempts, while the trained policy outperforms four frontier models in evidence‑based forecasting, achieving lower soft‑Brier scores at roughly 5% of the inference cost. The authors release the environment, dataset, and per‑rollout records as a reusable harness for temporal forecasting agents.
By Yusuf Afifi, Artur Kiulian, Anton Polishko, Mykola Khandoga, Hamudi Naanaa, Alina Krasnobrizha
The paper investigates the phenomenon of self‑repair in language models, proposing that it arises from a pre‑existing gain in components that act as counterweights when a component is ablated. By modeling interventions as points on a counterfactual axis, the authors derive an affine law for the causal repair response of fine‑grained units, showing that most downstream directions across several models follow this law. They further demonstrate that the slope of this law can be predicted from fixed weights, suggesting that self‑repair is a predictable, counterweight‑driven response rather than a noisy, unexplained effect.
By Areeb Ahmad, Pratinav Seth, Vinay Kumar Sankarapu
The paper evaluates Jev, a general-purpose decision model, for classifying network traffic using only the first ten packets’ sizes, directions, and inter‑packet times. Across 52,000 records, adding 40 labeled examples improves Jev’s accuracy from 9.80% to 28.42%, yet Random Forest and Extra Trees models trained on 8,000 records achieve higher accuracies of 69.95% and 66.80%. Even with 150 examples, Jev reaches only 34.50% accuracy, and in a paired 100‑record test, Jev’s 29% accuracy and 0.750 s median request time are lower than the generative OpenAI GPT‑5.6 Sol’s 37% accuracy and 6.036 s response time, though Jev incurs lower API charges.
By Shenghe Xu, Lifan Mei
The study evaluates 4‑bit quantization and low‑rank adapter fine‑tuning (QLoRA) on several large protein language models, finding that many model‑task pairs retain over 90% of full fine‑tuning performance while achieving up to 90% GPU memory savings. QLoRA preserves early‑layer representations and induces task‑specific changes in later layers, closely resembling full fine‑tuning with smaller representational shifts. For generative models, 4‑bit quantization largely maintains structural and sequence‑level properties, though token‑level analysis reveals model‑dependent changes in autoregressive output distributions.
By Ilan Yaniv Zeisler, Sebastian Clancy, Pouriya Bayat, Saaim Raad, Ivan Kraskov, Matthew Xie, Vivian White, Spencer Perkins, Serena Singh, Sepehr Bayat, Keith Pardee
SkillSpec is a two‑phase framework for evolving natural‑language skills in large language model agents. The first phase, consensus‑gated evolution, generates candidate skills from complementary editing intents and commits updates only when paired evaluations reach consensus on overall improvement and non‑negative aggregate gain. The second phase, representation specialization, uses signals from the optimization trajectory to choose an appropriate flat, graph, or hybrid structure for the skill, improving success rates by an average of 6.89% over SkillOpt across six benchmarks and three target language models.
By Huancheng Chen, Xiaodi Sun, Zhaoqiong Huang, Shenyang Huang Shreya Singhal, Jingwen Lu
The paper introduces entry-state sharpening, a data‑free pre‑training step that prepares a language model’s checkpoint in a sharper, lower‑entropy state before test‑time reinforcement learning (TTRL). By reducing policy entropy, the model can more efficiently use its limited adaptation budget, leading to higher endpoint conversion efficiency across tasks such as MATH, GPQA, and AMC. Experiments with different data‑free objectives (e.g., R‑Zero vs. SPIRAL) demonstrate that the choice of pre‑training objective strongly influences the checkpoint’s readiness for TTRL, and a label‑free self‑distillation intervention can further sharpen the entry state.
By Zhanming Zhang, Vinoth Selvendran
The paper introduces a supervised algebraic counting field (ACF) that predicts cryptic ligand‑binding pocket residues from apo protein structures by compiling geometric, physicochemical, and topological features into compact integer‑weighted lookup tables. ACF allows each prediction score to be reconstructed from feature values, training counts, table weights, and spatial aggregation, eliminating the need for sequence search, structural‑template transfer, or a protein language model during inference. In benchmark tests on CryptoBench and external datasets, ACF outperformed the P2Rank method in mean paired ROC‑AUC on one external set, though gains varied across datasets and were not universally confirmed.
By Shan Yu, Xuening Wu
ScaffoldM3C is a lightweight, multimodal, auto‑regressive framework that generates stable 3D block constructions by treating the task as a probabilistic next‑block generation problem. It incorporates text, image, and sketch conditioning, introduces a scaffold block token to aid intermediate stability, and uses Sequential Monte Carlo to explore multiple assembly sequences simultaneously. The model is four times smaller than existing baselines, achieving 5‑ to 20‑fold inference speedups while matching or surpassing state‑of‑the‑art construction quality and stability in both simulations and real‑world robot demonstrations.
By Gadiel Sznaier Camps, Chengyang He, Guillaume Sartoretti, Eduardo Montijano, Mac Schwager