Every paper checked — nothing hidden

All papers

Every paper checked in the week of 12–18 Oct 2026 · LLMs for code — nothing is hidden. Brief papers first.

All1,106 In brief3 More (on-topic)144 Not relevant928 Not assessed6 Older42 Could not confirm7 Blogs with no paper39 All blogs42

Paper	Topic	Relevance
Specification-Based Testing of Large Language Models The paper introduces a testing framework that turns documented model safety claims into specifications to find counterexamples via mutated prompts. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Applies software testing principles to evaluate LLMs, not code generation.
Scale, Drivers, and Limits of Representational Alignment Across Large Language Models The paper presents two geometric methods to evaluate where and why representational structures align across different large language models. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Not related to software engineering or code generation models.
What Counts as the Same Task? Executable Contracts for Robustness Evaluation of Tool-Using Agents The paper proposes executable contracts to verify that adversarial benchmark edits preserve the core task definition when evaluating tool-using agents. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Addresses agent task contracts generally rather than software code generation.
AphaVoice: Text-to-Speech Model for Aphasia Patient Simulation The paper develops a text-to-speech model that simulates realistic clinical dysfluency in aphasia speech across multiple speakers for training healthcare professionals. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Covers speech generation rather than software engineering or coding.
Decoupling Knowledge Forgetting and Refusal Learning in MLLM Unlearning The paper proposes a distillation framework to remove specific multimodal knowledge without causing hallucinations or leaving residual queryable information. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Does not address code assistants, programming benchmarks, or software repair.
Towards Merging-conditional Backdoor Unalignment against Large Language Models This paper demonstrates a security risk where dormant backdoors trigger safety failures only after merging models that passed individual safety audits. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Does not involve software engineering, code generation, or program repair.
Lingdan-V2: Large language models for clinically grounded reasoning in traditional Chinese medicine This paper introduces an aligned language model family trained with structured reinforcement learning for clinical reasoning and traditional Chinese medicine prescribing. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Does not involve code generation, software assistants, or program repair.
How to Build a Benchmark That Actually Measures Taste This paper introduces a pairwise evaluation benchmark and protocol designed to measure subjective aesthetic taste while controlling for shortcut artifacts. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Does not involve coding assistants, program repair, or software engineering.
Vela: User Behavior as Supervision Sparks LLM Intelligence This paper presents a framework that converts long-term real user behavior records into supervisory tasks to train personalization-focused language models. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Deals with user behavior modeling, not code generation or software engineering.
CUSE: Fine-Grained Multimodal Unlearning with Counterfactual Guidance and On-Policy Self-Distillation This paper introduces a multimodal unlearning method using counterfactual inputs and on-policy self-distillation to remove sensitive data without damaging general knowledge. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Deals with multimodal unlearning, not code generation or program repair.
AI-Native Product Management: A Vision This position paper outlines a conceptual operating model and set of principles for integrating persistent AI agents into product management. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Focuses on product management and organization rather than code generation or software repair.
WorkBorne: Discovering Agentic Tasks from Real-World Artifact Transitions This paper presents a framework for discovering and synthesizing terminal-based agent tasks grounded in real-world artifact transitions. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Focuses broadly on terminal-based agent tasks rather than code generation or repair.
SkillForgeBench: Evaluating Agent Reliability Grounded in Agent Skill Specifications This paper introduces a benchmark that uses agent skill specifications to evaluate the reliability and procedural correctness of AI agents. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant General agent workflow evaluation, not specialized software engineering or code generation.
Predicting how model propensities generalize out of distribution This paper studies how language model behavioral propensities generalize out-of-distribution across different elicitation methods. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Does not address code generation, program repair, or software engineering.
What Survives When Activation Steering Is Used as Memory This paper evaluates the capabilities and limitations of using activation steering vectors as an agent memory mechanism. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Studies memory storage in language models, not code generation or program repair.
UltraBERT: Beyond the Limit of BERT Pre-Training This paper analyzes pretraining factors such as masking ratios and token budgets to optimize bidirectional transformer encoders. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Deals with general language encoder pre-training, not software engineering models.
Toward System One Clinical Intelligence: Parameter-Efficient Fine-Tuning and Counterfactual Explainability of Decision Foundation Models This paper fine-tunes a quantized decision model to classify diseases from patient symptom descriptions without free-form text generation. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Addresses clinical text classification, not code generation or software engineering.
Select or Rewrite: Oracle-Guided Fine-Tuning for Generative ASR Post-Editing from N-Best Hypotheses This paper fine-tunes language models to selectively correct speech recognition transcripts using curated candidate pools. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Covers speech-to-text transcript correction, not software coding or program repair.
Have You Been Convinced? A Neuro-symbolic Argumentation Reasoning Approach to Detecting Persuasiveness in ChangeMyView Discussions This paper presents a neuro-symbolic framework using language models and formal argumentation semantics to detect persuasiveness in online discussions. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Focuses on discourse persuasion analysis, not software engineering or code generation.
UDHRBench: Benchmarking Human Rights in Large Language Models Evaluates commercial language models against Universal Declaration of Human Rights principles, revealing misalignments driven by efficiency arguments. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Assesses legal and ethical norms, not code synthesis or repair.
Exploring Jailbreak Vulnerability Across Depth-Based Dynamic Inference in Large Language Models Studies how depth-based early-exit inference affects language model safety, finding shallow exits resist jailbreaks better than deep exits. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Addresses model safety and inference depth, not code generation.
PANINI-DR: Controlled Deep Research through Generative Semantic Workspaces Presents a multi-agent framework that decouples research planning, subquestion investigation, and factual verification via a shared semantic workspace. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Addresses information search and research agents, not code generation or repair.
North or South? Diagnosing LLMs' Understanding of North Korean Language, Society, and Knowledge The paper presents LENS, a benchmark evaluating language models on North Korean linguistic nuances, sociocultural perspectives, and domain knowledge. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Assesses sociolinguistic and regional knowledge, not code generation or repair.
A Sovereign, Indic-First, Reasoning Model The paper presents a hybrid language model adapted for reasoning, instruction following, and tool use across English and eight Indian languages. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant General reasoning model rather than a specialized code assistant.
In-Place Indic Tokenizer Expansion and Continual Pretraining of Nemotron-3-Nano-30B-A3B This paper details vocabulary expansion and continual pretraining techniques to adapt an open language model across eight Indic languages. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Focuses on natural language pretraining rather than code generation.
Sensitivity or Criterion? Measuring and Moving When Language Models Act This paper introduces a diagnostic framework using signal detection theory to measure when language model agents decide to use tools. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Evaluates physical state tool actions, not software code generation.
Hidden in Plain Sight: Advancing Harmful Content Understanding via Perception-Aware Adversarial Self-Play This paper presents an adversarial self-play framework using multi-perceptual views to improve language model detection of obfuscated harmful text. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Does not address code generation or software engineering tasks.
When Messages Go Wrong: Understanding Communication Faults in LLM-Based Multi-Agent Systems This paper provides an empirical study and taxonomy of communication faults and error propagation across multi-agent language model systems. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Studies multi-agent communication networks rather than software coding models.
REPRO: Reflective Evolutionary Prompt Optimizer for Automated LLM Red-Teaming This paper proposes a black-box evolutionary prompt optimization framework using a small open-weight model to discover adversarial jailbreaks against language models. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Focuses on adversarial jailbreaks, not software engineering models.
SARATHI: Decoupled Data-Free Probing and Adaptive Reconstruction for Structured LLM Pruning This paper presents a structured pruning framework that decouples data-free neuron selection from localized weight reconstruction to reduce model memory requirements. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Compression method for LLMs, not code-specific software engineering.
bharatSetu: a 1M x 22 bridge for indic translation The paper introduces a synthetic parallel translation dataset covering 22 scheduled Indic languages aligned via English pivot sentences. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Deals with natural language translation, not programming or code generation.
Agent Sociodrama: Social-Dynamics-Guided Persona Recasting for LLM Agent Teams The paper introduces a framework to model evolving social dynamics and update personas in multi-agent language model teams. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Investigates social persona dynamics, not software engineering or code generation.
An Explainable System for Offensive Content Detection in Nastaliq Urdu: Corpus Curation, Transformer Benchmarking, and Word-Level Attribution The paper presents a Nastaliq Urdu comment corpus and evaluates transformer models with word-level attributions for offensive content detection. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Concerns natural language moderation, not coding or software engineering.
Adaptive Look, Flexible Thought: Learning How and How Long to Think in Multimodal Reasoning The paper proposes a reasoning framework that discretizes visual states into tokens to jointly optimize visual computation and chain-of-thought steps. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Deals with multimodal chain-of-thought, not code generation or program repair.
Personalized LLM Recommendation with Semantic IDs: Autoregressive Beam Search vs. Multi-Token Prediction The paper compares single-head autoregressive beam search with parallel multi-token prediction for generating semantic IDs in recommendation systems. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Deals with recommendation systems, not software code generation or repair.
Review Rarely Beats the Stronger Model: Separating Verification from Capability Transfer in LLM Pipelines The paper shows that LLM review pipelines rarely outperform the stronger model alone due to confirmation bias rather than true verification. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Tests general verification pipelines across tasks, not specifically code generation systems.
The Anatomy of a Blind Spot: What an LLM Judge Must Be Shown, and How Much The paper tests what context judge models require, showing that omitting source constraints degrades the quality of iterative feedback loops. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Focuses on general judge evaluation setups, not coding tools.
Unfair and Unsafe: Everyday Bias Triggers Emergent Misalignment The paper finds that fine-tuning language models on polite text containing demographic stereotypes triggers broader safety and truthfulness failures on unrelated tasks. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Does not address code generation, program repair, or software engineering.
Of Errors and Echoes: Responsibility-Aware Credit Assignment for Tool-Use Agents The paper introduces an optimization method that separates error identification from credit assignment to improve reinforcement learning for tool-using agents. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Evaluates tool-use credit assignment rather than software development assistants.
When Agents Talk, Secrets Travel: Topology-Driven Leakage and Recipient-Conditioned Sanitization in Multi-Agent LLM Systems The paper investigates privacy leakage caused by multi-agent network topologies and proposes an execution sanitizer conditioned on recipient identity. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Examines multi-agent privacy topologies, not software engineering code generation.
When Agent Evaluation Becomes Unreliable: An Empirical Study of Oracle Failures and Configuration-Dependent Measurement in an LLM Agent Benchmark The paper audits agent evaluation benchmarks for software defects, showing that broken reward functions and configuration shifts substantially distort reported success rates. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Focuses on agent test harness integrity rather than code synthesis assistants.
Dream2Discern: Learning from Hard Latent Variants for Runtime Failure Prediction in Multi-Agent Systems The paper presents a runtime auditor that predicts execution failures in multi-agent systems using loss-guided perturbations to identify hard failure cases. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Targets multi-agent runtime auditing, not software engineering code generation.
GUIDE: Computationally Augmented Consensus for Trustworthy Clinical Practice Guidelines This paper introduces GUIDE, a multi-agent copilot system that assists clinicians in drafting and updating evidence-based medical practice guidelines. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Not related to coding assistants or program repair.
PID-Gated Continual Prompt Adaptation Under Noisy Feedback This paper proposes control-inspired gates to decide when language model agents should adapt their prompts without absorbing noisy user feedback. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Does not address code generation or software engineering.
Query Your Way Out: Benchmarking Multi-Turn Knowledge Reasoning through Generalized Puzzle Solving This paper introduces the Active Probing Benchmark to evaluate multi-turn active questioning and specialized domain reasoning in language models. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Not focused on code generation or software engineering.
Omni-Modal LLMs Know More Than They Say This paper analyzes why omni-modal language models perceive non-textual inputs correctly yet fail to mention them in text outputs. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Not related to programming, code generation, or software engineering.
Phantom Merge: When Your Large Language Model Agents Pick One but Tell You About Another This paper identifies and mitigates Phantom Merge, an error where language model agents bind retrieved facts to the wrong entity. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Not focused on code generation or software engineering tasks.
SafeDSE: Joint Exploration of Safety Architecture Patterns, Mapping, and Scheduling for Dataflow Applications With Safety Arguments (Extended Version) Presents a design space exploration tool for mapping, scheduling, and generating safety arguments for dataflow applications on embedded hardware platforms. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Not related to LLMs writing or fixing code.
Avoiding What Your Teacher Already Knows: Entity Familiarity for Deep-Search Distillation Measures teacher entity familiarity to filter out uninformative search trajectories when distilling deep-search capabilities into smaller agents. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Not focused on coding assistants or software engineering.
agent-risk: Risk Assessment for Changes to AI Agents Provides a framework for assessing risks of proposed changes to deployed AI agents by selectively replaying execution logs and saved states. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Focuses on agent policy replay, not software development code.
REVEAL: Reverse Engineering Agentic Latents from Frontier-Model Traces Extracts reusable local control signals from frontier model traces to guide smaller language model agents without extra target-model training. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Not focused on code generation or software engineering tasks.
RetroPlanner: Multi-Route Expansion and Multi-Objective Reasoning for LLM-Based Retrosynthesis Planning Proposes a multi-route retrosynthesis planner that evaluates synthesizability and cost across chemical reaction options. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Applies to chemistry synthesis planning, not software code generation.
ARMED: Adaptive Retrieval-Augmented Medication Recommendation Proposes an adaptive retrieval-augmented generation framework that dynamically expands clinical queries for medication recommendation. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Addresses medication recommendation, not software coding or program repair.
VisExMEM: Visual Evidence-Based Explainable Multimodal Evaluation Metric Presents a reference-free evaluation metric that decomposes captions into atomic claims and verifies them against visual evidence. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Addresses image caption evaluation, not code generation or review.
PATMT2: Two-stage domain adaptation for multilingual patent translation Evaluates a two-stage domain adaptation method for multilingual patent translation and analyzes failure modes in low-resource languages. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Covers patent text translation, not code generation or program repair.
EAU: Environment-Agent-User Co-evolution for Training User-Centric Agents Co-evolves the user simulator, environment tools, and rewards during training to improve interactive multi-turn language model agents. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Addresses general user-interaction agent training, not software engineering models.
PatTransBench: an ipc-stratified, multi-dimensional benchmark for patent machine translation evaluation Presents an evaluation benchmark and multi-dimensional protocol for patent machine translation across Chinese, English, Japanese, and Korean. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Focuses on patent language translation rather than software code generation.
MixLID: Token-Level Language Identification in Mixed-Language Text Classifies languages at the token level across 111 language-script combinations in mixed-language and code-switched texts. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Applies to multilingual human text rather than software code or coding assistants.
SuPER-Mamba: Surprisal Prioritization for Extended-Context Routing in Mamba Improves long-context handling in Mamba state space models without retraining by using surprisal scores to guide state routing. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Evaluates general long-context benchmarks rather than code generation or repair.
Does a Hosted Decision Model Respect Irrelevant Alternatives? A Pre-Registered Test of Option-Set Dependence in Jev Tests a hosted decision model to show that adding irrelevant options systematically shifts output probability distributions between top candidates. Preprint	LLMs for code via arxiv · 15 Oct, 06:17	Not relevant Evaluates classification decision models rather than code generation or repair.

781–840 of 1,106← Previous Next →