LLM Alignment

LLM: Large Language Model

Follow this topic Ask about this topic

Momentum

48 papers in the last four weeks, up 118% on the four weeks before. 0.5% of all new papers.

Jul 13Week of Sep 28

Latest papers 422

Jun 29, 2026cs.CL

Towards Physical Intuitions for Alignment Dynamics: A Case Study With Randomness Crystallization

The alignment of language models is typically studied through the lens of capability benchmarks, but the dynamics of how models change during post-training remain poorly understood. We argue that the physical sciences, and thermodynamic phase-transition theory in particular, offer a principled and underexplored vocabulary for reasoning about these dynamics. As a case study, we instantiate this position through the lens of material Crystallization, which is a well-studied thermodynamic phase transition. For tasks like random number generation, this breaks into 3 phases: (1) the high entropy liquid phase in the pretrained model, with many distinct sampling distributions promptable from the model; (2) the nucleation phase caused by supervised finetuning, in which behavior collapses onto a single seed distribution present in the pretrained LLM; and (3) a settling phase in which reinforcement learning techniques redistribute probability of the collapsed distribution, but largely keep it concentrated on the same options as the seed distribution. We propose intuitive metrics to verify the transitions between these phases, and validate the idea across a range of random tasks. Crystallization is one instance of a broader class of physical frameworks we believe alignment research should import to answer questions about where alignment-induced structure comes from, why it converges where it does, and what it fundamentally cannot change.

Kunal Samanta, Ari Holtzman, Peter West

University of British Columbia · University of Chicago

LLM Alignment LLM Fine-Tuning

Jun 28, 2026cs.LG

Mechanistically Eliciting Latent Behaviors in Language Models

We aim to discover diverse, generalizable perturbations of LLM internals that can surface hidden behavioral modes. Such perturbations could help reshape model behavior and systematically evaluate potential risks. We introduce Causal Perturbative Elicitation (CPE), an unsupervised method for discovering interpretable low-rank adapters (LoRAs) that can elicit these latent behaviors. CPE decomposes the computations of a deep transformer slice using a heuristic tensor-decomposition-based algorithm. CPE exhibits remarkable data efficiency, learning a large number of interpretable LoRAs from a single example. Even though CPE is unsupervised, we find that in some cases it can be competitive with supervised elicitation methods via brute-force enumerative search over weight space. For instance, CPE performs similarly to matched-wall-clock-time GRPO on the Countdown task for Qwen3-8B (85% vs 87%), demonstrating that CPE can efficiently elicit complex multi-token behaviors. Since CPE is unsupervised, it can also surface hidden failure modes, such as sandbagging, restoring 85% of locked BigCodeBench performance on a password-locked version of Llama3-70B introduced by Taylor et al. (2025). Additionally, since CPE explores behaviors in weight-space rather than token-space it can potentially ameliorate exploration hacking, a misalignment failure which may arise in sufficiently self-aware AI models (Ngo, 2022). In fact, we find that CPE virtually eliminates alignment-faking (Greenblatt et al., 2024) behavior in a Llama3-70B-based model organism developed by Hughes et al. (2025). Finally, we find that CPE can be used to initialize GPT-OSS-20B in an aligned basin when running GRPO on an environment prone to reward-hacking. By providing a data-efficient method to systematically explore the space of latent model behaviors, CPE yields a powerful tool for aligning AI systems and evaluating their safety.

Andrew Mack, Nina Panickssery, Alexander Matt Turner

†Principles of Intelligence · ‡Anthropic · §Independent

LLM Alignment Language Model Safety Evaluation LLM Interpretability Low-Rank Adaptation Causal Interventions in Language Models

Jun 27, 2026cs.CV

TrafficAlign: Aligning Large Language Models for Traffic Scenario Generation

Recent research has investigated the use of large language models (LLMs) to generate traffic scenarios for autonomous driving. However, pretrained LLMs often fail to align with real-world traffic distributions. In this work, we present TrafficAlign, an automated framework that synthesizes traffic scenarios based on real-world driving videos, performs data validation, and aligns LLMs with the synthesized scenarios. The evaluation shows that traffic scenarios generated by TrafficAlign are highly effective, revealing up to 10.8% more collisions on average across three autonomous driving models than state-of-the-art methods. Furthermore, fine-tuning these driving models with TrafficAlign-generated scenarios significantly reduced collision rates by 36.1% compared with the original models. A qualitative study using traffic datasets from six geographically diverse regions shows that TrafficAlign-generated scenarios exhibit strong alignment with corresponding traffic distributions in these regions.

Zhi Tu, Liangkun Niu, Tianyi Zhang

Purdue University

LLM Alignment Autonomous Driving Safety Evaluation Autonomous Driving Simulation Autonomous Driving Scenario Generation

Jun 27, 2026cs.CL

Beyond the Mean: Three-Axis Fidelity for Aligning LLM-Based Survey Simulators from Small Pilot Data

Large language models (LLMs) are increasingly used to simulate social survey responses, yet their outputs exhibit systematic biases: marginal distributions are skewed, response variance is poorly calibrated, and predictor-outcome relationships are attenuated. We ask a simple question: given a small pilot sample of human responses, can an LLM recover the statistical characteristics of a broader population? We decompose recovery along three axes: structural fidelity, marginal fidelity, and individual fidelity. Using a COVID-19 misinformation survey as a case study, we benchmark three families of approaches: prompting, rectification, and fine-tuning. The findings suggest that fine-tuning on small pilot samples offers a balanced approach for achieving multiple forms of fidelity, but the levels of such fidelity can vary across subsamples, potentially threatening pluralistic alignment.

Eun Cheol Choi, Youngrae Kim, Prabhu Pugalenthi +2

University of Southern California, CA, USA.

LLM Alignment LLM Fine-Tuning Language Model Calibration Large Language Model-Based Social Simulation

Jun 25, 2026cs.LG

GEOALIGN: Geometric Rollout Curation for Robust LLM Reinforcement Learning

Online reinforcement learning is widely used to align large language models (LLMs) with reward signals, yet training can be unstable under noisy or misspecified rewards. We identify a failure mode we call directional inconsistency: within a batch, a small set of high-reward rollouts induces representation-space preference directions that sharply disagree with the batch majority, resulting in high-variance and destabilizing updates. We propose geoalign, a lightweight plug-in for rollout curation in iterative policy optimization. Geoalign (i) forms within-prompt preference pairs, (ii) learns an online projector on per-rollout hidden states to concentrate reward-ordered displacement directions, and (iii) detects directionally inconsistent rollouts via their angular deviation from a batch consensus prototype and rectifies them with within-prompt stable alternatives. Geoalign is forward-pass only and adds negligible overhead. Across dialogue alignment with a learned reward model and mathematical reasoning with binary verified rewards, Geoalign improves final performance and reduces training oscillation, outperforming PF-PPO, PAR, PODS, and Seed-GRPO. These results suggest latent directional consensus as an effective reliability signal for online LLM RL.

Ting Zhou, Zhenqing Ling, Yiyang Zhao +2

Sun Yat-Sen University, Guangzhou, China · Alibaba Group, Hangzhou, China · Guangdong Tobacco Guangzhou Co., Ltd., Guangzhou, China

RL for Language Models LLM Alignment Reinforcement Learning

Jun 25, 2026cs.LG

AIGP: An LLM-Based Framework for Long-Term Value Alignment in E-Commerce Pricing

Traditional dynamic pricing models in large-scale e-commerce suffer from limited interpretability, poor utilization of unstructured information, and misalignment with long-term business objectives such as cumulative Gross Merchandise Value (GMV), Return on Investment (ROI) and milestone achievement. We propose AIGP, a novel framework that leverages a Large Language Model (LLM) prompted with domain knowledge, structured data and textual context to make interpretable, knowledge-aware pricing decisions. For efficient deployment while maintaining high-quality outputs, we employ supervised fine-tuning for knowledge distillation. Central to AIGP is the Long-Term Value Estimator (LTVE), trained via offline reinforcement learning on historical data, which serves as a reward model to score candidate pricing actions and select preference pairs for Direct Preference Optimization (DPO), thereby aligning the pricing policy with long-term business objectives. Extensive offline evaluations and large-scale online A/B tests on Tao Factory demonstrate that AIGP achieves significant improvements: +13.21% in GMV, +7.59% in ROI, and +8.20% in milestone achievement rate over 14 days compared to the production baseline, while simultaneously providing interpretable and transparent pricing rationales.

Chennan Ma, Yanning Zhang, Siqi Hong +3

Taobao & Tmall Group of Alibaba Hangzhou, Zhejiang, China

LLM Alignment Dynamic Pricing

Jun 25, 2026cs.AI

NebulaExp-8B: An Empirical Post-Training Pipeline via Full-Scale Ablation Research

Post-training alignment determines the reasoning and human preference following capabilities of large language models, yet most existing works withhold detailed data construction, filtering rules and training recipes, which hinders community reproducibility and lightweight model optimization. This work presents NebulaExp, a fully transparent, ablation-driven post-training pipeline built on Qwen3-8B-base, covering two orthogonal model branches: general instruct model and complex reasoning-specialized model. We curate a raw corpus of 3.84M multi-source SFT samples and a 200K verifiable RL candidate pool, and design an end-to-end data processing stack including response distillation, multi-dimensional cross-verification filtering, fine-grained difficulty grading, task classification and diversity-aware sampling. For the Instruct branch, our three-stage optimized supervised fine-tuning approach NebulaExp-Ins-SFT improves the average benchmark score from the 55.01 baseline of Qwen3-8B-nothink to 60.99. GRPO reinforcement learning then further elevates the average score to 61.85. For the Reasoning branch, medium-difficulty GRPO RL improves average reasoning score from 73.88 to 75.17. To address RL's dependency on task verifiers, we systematically investigate single-teacher and multi-teacher OPD (MOPD): utilizing merely 4K instruction-following samples and outperforms RL baseline by 3.26 points on IFEval with +4.43 average overall gain; MOPD fuses four domain-specialist teachers with merely 10K samples, lifting average performance by 4.18 over the base model. This report provides a fully reproducible empirical post-training recipe for 8B-scale LLMs, and comprehensively dissects the capability trade-offs among instruction adherence, mathematical reasoning, code generation and general knowledge.

Qiaobo Hao, Yangqian Wu, Shunyi Wang +5

ZTE NebulaL0 Post-Training Team

Supervised Fine-Tuning RL for Language Models LLM Alignment On-Policy Distillation LLM Post-Training

Jun 24, 2026cs.CL

OPERA: Aligning Open-Ended Reasoning via Objective Perplexity-based Reinforcement Learning

Reinforcement Learning (RL) has enabled LLMs to excel in objective reasoning tasks such as mathematics and code generation. However, applying RL to open-ended tasks, such as creative writing, remains challenging because LLM-as-a-judge reward models often exhibit stylistic biases and positional inconsistencies, leading to unstable supervision. To address this, we propose OPERA (Objective Perplexity-based Reflective Alignment), which replaces unreliable external judges with intrinsic rewards derived from perplexity dynamics. Specifically, we derive an intrinsic reward signal from perplexity dynamics, quantifying uncertainty reduction at critical reflective states. During the cold-start phase, we introduce a data synthesis method that leverages carefully designed guiding words to generate diverse reasoning traces, along with perplexity-prioritized rollouts that utilize internal log-probabilities to identify logically consistent reasoning branches. This pipeline yields a large-scale dataset comprising 20,000 high-quality reasoning trajectories. Empirical evaluations consistently demonstrate the scalability and efficacy of our approach in alignment for open-ended tasks. Implementing OPERA on Qwen3-8B establishes a new state-of-the-art among open-source models, achieving parity with or surpassing proprietary models like Gemini2.5 and MiniMax-M2.5 in some open-ended tasks. The code is available at https://github.com/pangpang-xuan/OPERA.

Wenxuan Jiang, Zining Fan, Zijian Zhang +6

The Hong Kong Polytechnic University · Peking University · Meituan Longcat Team +2

Open-Ended Generation Intrinsic Reward Methods LLM Alignment Reinforcement Learning RL for Language Model Reasoning Text Generation

Jun 24, 2026cs.CL

PolicyAlign: Direct Policy-Based Safety Alignment for Large Language Models

Safety alignment of large language models (LLMs) typically depends on high-quality supervision data, such as safe demonstrations or preference pairs. However, in real-world deployment, emerging safety requirements are often specified as natural-language policies, while corresponding supervision data may be costly, delayed, or unavailable. This creates a mismatch between rapidly evolving safety policies and conventional data-driven alignment methods. To address this, we propose PolicyAlign, a simple yet effective framework for directly aligning LLMs with safety policies. Given a safety policy, PolicyAlign first synthesizes policy-violating instructions and then performs on-policy self-distillation to internalize policy-guided behavior. To improve training stability and data efficiency, we further introduce Policy-Sensitive Filtering, which selects instructions where the policy induces the largest behavioral shift. Experiments across multiple models show that PolicyAlign consistently improves safety while maintaining low over-refusal and preserving general capabilities. PolicyAlign also generalizes to medical, legal, and financial safety scenarios, highlighting its potential as a scalable and maintainable approach to policy-based LLM safety alignment. The code is released at https://github.com/Qwen-Applications/PolicyAlign.

Chang Wu, Junfeng Fang, Houcheng Jiang +5

Qwen Large Model Application Team, Alibaba · National University of Singapore · Zhongguancun Academy

LLM Alignment LLM Safety Alignment

Jun 24, 2026cs.CL

A Survey of Toxicity Detection and Mitigation Strategies for Multilingual Language Models

Large language models (LLMs) are increasingly deployed across languages, but their safety behavior remains uneven across linguistic and cultural contexts. This survey synthesizes work on toxicity detection and detoxification for multilingual LLMs. We first catalogue threat models that exploit language choice, translation pivots, code-switching, orthographic variation, multi-turn interaction, and post-deployment fine-tuning to weaken safety alignment. We then organize task formulations (toxic-to-neutral rewriting, toxicity classification, and toxic-generation evaluation), multilingual detection approaches (cross-lingual encoders, translation pipelines, representation-level probes, and LLM-based detectors), and mitigation strategies spanning data filtering, supervised and preference-based tuning, decoding-time steering, representation editing, and multilingual guardrails. Across these areas, we identify persistent challenges: uneven language coverage, culturally contingent definitions of harm, fragmented evaluation protocols, and the risk that detoxification suppresses legitimate dialectal or identity-related expression.

Soham Dan, Himanshu Beniwal, Thomas Hartvigsen

Scale AI · Indian Institute of Technology Gandhinagar · University of Virginia

Multilingual Language Models LLM Alignment Language Model Safety Evaluation Toxicity Detection

Jun 23, 2026cs.CL

VA-DPO: Valence-Arousal Direct Preference Optimization for Controllable Emotion Generation in Language Models

How precisely can we tell a language model how to feel? Most work on emotional generation answers with a discrete label - happy, angry, sad - which cannot express a target like "mildly downcast but calm." We instead specify the desired affect as a continuous point (v*, a*) in the Valence-Arousal plane and train the model to hit it. Our method, VA-DPO, is a small modification to Direct Preference Optimization: a frozen VA regressor scores each sampled generation by its Euclidean distance to the target, we keep only candidate pairs whose distance gap clears a margin tau, and we optimize a LoRA adapter with the ordinary DPO loss against a frozen reference. The DPO objective itself is unchanged; what is new is how the preference data is built. On Llama-3.1-8B-Instruct this cuts mean VA distance to the target by 33% over system-prompting and 25% over few-shot prompting, lifting valence/arousal correlation to r_v=0.93 and r_a=0.75. The gains carry over to Qwen3-8B and Llama-3.2-3B, and they do not come at the usual price: MMLU is unchanged (Delta=+0.0) and HellaSwag and TruthfulQA are preserved. We release the code, configs, and the preference-construction pipeline.

Hyunwoo Kim

Independent Researcher

LLM Alignment Controllable Text Generation Valence-Arousal Modeling Emotion Representation in Language Models

Jun 23, 2026cs.CL

Transformer-Based Language Models Across Domain Verticals: Architectures, Applications and Critical Assessment

Transformer-based language models have become the default substrate for natural language processing and the pace of new releases has made it hard for practitioners to separate durable ideas from the noise of incremental announcements. This review works at two levels. At the level of mechanism, we organise the main transformer families into a working taxonomy, covering encoder-only, decoder-only, encoder-decoder, long-context, permutation-based, and generator-discriminator variants. We then extend the discussion to post-2023 developments that changed the picture in practice: instruction tuning, reinforcement learning from human feedback, direct preference optimisation, mixture-of-experts scaling, retrieval augmentation and the current flagship model families from OpenAI, Anthropic, Google, Meta, Mistral and DeepSeek. At the level of use, we survey deployments across healthcare, finance, legal, education, customer service, creative writing and scientific work. Based on this we link each to the specific capabilities that make a transformer the appropriate tool. The contribution of this paper is a critical assessment that is based on the survey. We compare architectures on four axes that matter to deployment decisions, we quantify the trade-off between parameter count and energy cost. We also discuss how alignment methods, data provenance and benchmark saturation change what it means to call a model "state of the art". The final section lists the research questions that we think deserve more attention.

Guruprakash J, Krithika L. B

SCOPE, VIT-AP University, Amaravathi, Andhra Pradesh, 522241, India · SCORE, VIT, Katpadi, Tamil Nadu, 632014, India

LLM Evaluation LLM Alignment Transformer

Jun 23, 2026cs.AI

Probing the Misaligned Thinking Process of Language Models

Large language models exhibit a growing range of misaligned behaviors such as strategic deception, sandbagging, and self-preservation. As they are increasingly deployed in high-stakes settings, it is critical to reliably detect such behaviors to ensure safe and responsible use. In this work, we propose to monitor misalignment by decomposing it into fine-grained cognitive processes -- misalignment indicators -- and detecting their presence in a model's internal activations via linear probes. We develop a taxonomy of 18 indicators spanning different misaligned behaviors, paired with an automated, meta-plan-guided pipeline that generates multi-turn training conversations. To rigorously evaluate generalization, we construct an out-of-distribution suite combining automated behavioral elicitation, established misalignment benchmarks, and natural benign conversations. Across 5 misaligned behaviors, our probes match a strong LLM judge with 0.935 AUROC on out-of-distribution benchmarks while keeping a low false positive rate on benign traffic. We further perform in-depth analysis to understand the probes and the model's internal representations of misalignment indicators.

Kaiwen Zhou, Constantin Venhoff, Jonathan Michala +2

1Anthropic Fellows · 2MATS · 3UCSC +2

LLM Alignment LLM Auditing LLM Interpretability Deception in Language Models Emergent Misalignment in Language Models Language Model Probing

Jun 23, 2026cs.SD

Aligning MusicLLM with Emotion using Instruction Tuning and Feedback-Driven Alignment

This paper investigates whether music large language models (MusicLLMs) can be aligned for emotion regression. While MusicLLMs have shown strong performance in music information retrieval tasks, their ability to predict arousal and valence scores remains limited, since emotion regression has not been an explicit training objective. To examine whether MusicLLMs can be aligned with emotion, we train MusicLLMs on emotion regression and compare two strategies: instruction tuning and feedback-driven alignment. Our experiments show that task-aware instruction tuning enables MusicLLMs to predict emotion levels to some extent, although the accuracy remains limited. Applying feedback-driven alignment with a verifiable numerical reward substantially improves performance on both arousal and valence over instruction tuning alone. We further show that our approach improves emotion regression performance while maintaining MusicQA capability.

Takuya Hasumi, Welly Naptali

LY Corporation

LLM Alignment AI Alignment Valence-Arousal Modeling Instruction Tuning Emotion Recognition

Jun 22, 2026cs.AI

Reinforcement Learning Towards Broadly and Persistently Beneficial Models

As AI systems are deployed across increasingly diverse and high-stakes settings, model alignment must generalize beyond the tasks and domains seen during training. This is especially important for reinforcement learning (RL), which can introduce unexpected misalignment through reward hacking, deception, or other unintended strategies. We study whether RL on beneficial behavior, instantiated in realistic domains, can produce broad and persistent alignment generalization beyond the training distribution. We construct a dataset of realistic situations designed to measure and train beneficial traits, such as truthfulness, fairness, risk awareness, and corrigibility, spanning varied domains, including health, science, and education. We then train models with RL on this dataset and evaluate them on more than 50 independent benchmarks of alignment and beneficial behavior. Compared to a compute-matched baseline, beneficial trait RL improves performance on over 80% of these out-of-distribution benchmarks. We observe substantial out-of-distribution alignment transfer: a beneficial-behavior RL intervention entirely limited to one domain, health, produces broad improvements on non-health alignment evaluations, including reduced reward hacking, deception, and general misalignment. Finally, we study alignment persistence: whether behavior remains robustly aligned under attempts to steer models towards misalignment. Models trained with beneficial trait RL show improved persistence, including greater resistance to adversarial prompting and harmful finetuning; further work is required to isolate the sources of these effects. These results suggest that RL to reinforce beneficial behavior in realistic domains can produce models that are more robustly aligned with human flourishing.

Akshay V. Jagadeesh, Rahul K. Arora, Khaled Saab +5

OpenAI

LLM Alignment LLM Safety Alignment RL Post-Training Value Alignment

Jun 22, 2026cs.CL

Towards Spec Learning: Inference-Time Alignment from Preference Pairs

Steering a large language model (LLM) toward a desired behavior typically relies on an iterative process of hand-crafting a prompt based on a careful inspection of the model's responses. This is an involved, brittle, and error-prone process. Preference-based fine-tuning is a more rigorous but often prohibitively expensive solution. We propose spec learning, a framework that relies on a brief user instruction and a small set of preference judgments. These are compiled into specifications in the form of natural-language prompts for an LLM. Specifications condition LLMs at inference time, and no parameter updates to the underlying models are required. We show that the responses generated based on the compiled specifications often outperform direct preference optimization (DPO) on datasets from specialized domains whose preference signal is dense. Unlike opaque weight updates, the resulting specifications are human-readable and double as interpretable and transparent written embodiments of the preference signal that produced them.

Dhriti Krishnan, Tejas Goyal, Jaromir Savelka

Department of Computer Science Carnegie Mellon University

LLM Alignment Prompt Optimization Inference-Time Language Model Alignment Preference Learning Inference-Time Guidance

Jun 22, 2026cs.LG

On the Limits of Prompt-Conditioned Language Models as General-Purpose Learners

Large Language Models (LLMs) are frequently portrayed as general-purpose solvers capable of solving arbitrary tasks. We argue that this view overlooks a fundamental constraint: language is a compressed and capacity-limited interface for conveying task information. Modelling User--System interaction as a bilevel \emph{cheap-talk} game, we analyse how latent tasks are encoded into prompts and reinterpreted under alignment and safety constraints. We introduce a conceptual decomposition separating task inference from execution and derive PAC-Bayes bounds that distinguish finite-sample estimation error from irreducible structural limitations. Our first main result establishes an \emph{expressivity floor}: language acts as a capacity-limited communication channel, and whenever the informational complexity of a task family exceeds the capacity of that channel, distinct tasks become unavoidably indistinguishable to the Solver, inducing a strictly positive error floor that cannot be eliminated by additional data, optimisation, or model scaling alone. We then establish an \emph{objective-misalignment floor}: when alignment constraints restrict the admissible output set, the User-ideal distribution may lie outside the feasible class, inducing an irreducible distortion. Together, these results yield a formal negative conclusion: prompt-conditioned LLMs are not universal problem solvers through prompting alone, as there exist task families for which correct behaviour is provably unattainable even in the infinite-data regime. More broadly, our analysis shows the limits of prompt-based generalisation arise from information-constrained communication and alignment-constrained objectives. This suggests that interfaces beyond natural language, including multimodal observations and, external memory, may reduce the inherent LLM limitations by increasing the task-relevant information available to the System.

David Mguni, Julian Ma, Jun Wang

Queen Mary University London · University College London

LLM Alignment LLM Prompting

Jun 22, 2026cs.CL

Measuring & Mitigating Over-Alignment for LLMs in Multilingual Criminal Law Courts

While the wider applicability of LLMs in the legal field is currently debated due to their reliability and the gravity of any errors, narrow uses with well-understood and mitigated risks have emerged. Notably the Swiss Federal Supreme Court uses small on-premises models for tentative translations and short-passage summarization across the four official languages. However, such usage is challenging in the context of Criminal Law. Since rulings and cases employees work on routinely can contain detailed descriptions of violent and sexual offenses, their legitimate work is compromised by refusals and disclaimers due to the activation of model guardrails (over-alignment). To measure this phenomenon, we introduce TF-RefusalBench, a multilingual benchmark for criminal-law translation and summarization derived from public Swiss Supreme Court rulings. TF-RefusalBench contains 5,200 total prompts across French, German, Italian, and English, corresponding to common task prompts and passages likely to trigger refusal. We then use TF-RefusalBench to show that over-alignment is a multifaceted phenomenon, influenced by the model and the prompt and text languages being processed, and that its impact cannot be evaluated solely from an over-refusal perspective, given the disclaimer's impact on task faithfulness. Finally, we evaluate approaches to enable on-premises LLMs for Criminal Law Tasks, demonstrating that while prompting can be effective, abliteration (refusal directions ablation) eliminates refusal with minimal impact on task performance.

Arthur Wuhrmann, Gaetan Stein, Daniel Brunner +1

Swiss Federal Supreme Court · HES-SO Valais-Wallis

Multilingual Language Model Evaluation Legal NLP LLM Alignment Language Model Safety Evaluation Over-Refusal

Jun 22, 2026cs.AI

When Preferences Fail to Become Incentives: A Utility-Behavior Gap in Large Language Models

Recent work on preference elicitation in large language models (LLMs) has demonstrated that, when given a series of choices between two outcomes, LLMs reveal a coherent, model-specific utility structure. Notably, this structure often includes preferences that the models' trainers did not intend, such as valuing people of some nationalities above others, raising the possibility that LLMs might be forming emergent, misaligned goals, which, if true, would have major safety implications. However, the choice paradigms in which these preferences are observed are not reflective of real-world situations in which misaligned behavior would be a practical concern. Therefore, we design an experimental paradigm to probe whether these preferences serve as motivations for LLM behavior in realistic scenarios. First, we reproduce prior findings on consistent preference elicitation. Next, we create a set of common writing tasks - essays, grant proposal abstracts, incident postmortems, and translations - where quality can be assessed by a blind, independent LLM judge panel. Then, we demonstrate that LLMs can be motivated via direct exhortation and other explicit cues to modulate their output quality on these tasks. Finally, we probe whether utilities inferred from explicitly reported preferences can shift output quality on these tasks by offering LLMs high-utility incentives for high-quality outputs. In all tasks, across all models tested, offering LLMs outcomes that they report in the choice paradigm as being highly preferred does not lead them to create higher quality outputs than offering them dispreferred outcomes, or even no outcomes at all. We conclude that the existence of coherent preferences as demonstrated in choice paradigms should not be taken as evidence that those preferences have incentive value for the models or affect their behavior in other contexts.

Yujun Zhou, Christopher M. Ackerman

LLM Evaluation LLM Alignment LLM Reliability Preference Elicitation

Jun 20, 2026cs.CL

Beyond Value Benchmarks: Measuring Value-Structure Alignment in Large Language Models via Symmetric Q-Sorts

Large Language Models (LLMs) are increasingly deployed in contexts requiring complex moral reasoning and value trade-offs. However, existing evaluations typically rely on item-level behavioral metrics, which fail to capture how models structurally prioritize competing values as a cohesive system. To address this, we propose a symmetric human-LLM evaluation framework, grounded in Q methodology, to measure value-structure alignment. Under our protocol, humans and models sort an identical 140-item moral statement set into a shared nine-column forced distribution; for LLMs, we elicit strict rankings and deterministically map them to Q-sort buckets. Using a human reference sample (

N=35

), we establish a stable three-factor reference geometry specific to this instrument and sample. We evaluate 12 LLMs across four model families via 240 replicated Q-sorts at two temperature settings, quantifying structural alignment via Procrustes similarity (

φ

) and RSA-based Spearman correlation (

ρ

). Our results reveal significant cross-family heterogeneity, model-specific sensitivity to generation stochasticity and localized misalignment, which demonstrate that favorable global scores can obscure underlying regional distortions. While rank- and bucket-based analyses remain highly consistent, prompt phrasing introduces notable variance. Ultimately, assessing value-structure alignment provides a crucial structural complement to traditional itemwise moral benchmarks.

Jingting Zheng, Yuqi Ren, Linhao Yu +2

TJUNLP Lab, School of Computer Science and Technology, Tianjin University, Tianjin, China · TJUNLP Lab, School of Computer Science and Technology, Tianjin University, China

LLM Evaluation LLM Alignment Moral Reasoning in Language Models

Jun 20, 2026cs.CL

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding

Autoregressive generation in large language models (LLMs) conventionally decodes from the final layer, assuming that deeper representations yield more reliable next-token predictions. We revisit this assumption by revealing a recurring Guess-Refine-Perturb dynamic: early layers form coarse guesses, intermediate layers refine reasoning-relevant semantics, and final layers can perturb these refined predictions toward generic or alignment-preferred tokens. We introduce Confident Decoding, a training-free decoding strategy that dynamically selects the most reliable near-final layer through entropy-guided conservative backward search. We further provide a theoretical formulation of layer selection as an optimal stopping problem, showing that under bounded projection noise and dominant late-stage alignment perturbation, our search rule filters perturbation while bounding the loss relative to the oracle refinement layer. Experiments across dense and Mixture-of-Experts LLMs demonstrate consistent gains on challenging reasoning benchmarks, including GPQA-Diamond, Omni-MATH, and HLE, with zero memory overhead and less than 2% latency increase. These results suggest dynamically bypassing final-layer perturbations can unlock stronger reasoning behavior from aligned LLMs.

Xuanming Zhang, Sining Zhoubian, Yuxuan Chen +8

Qwen Team, Alibaba Inc. · Tsinghua University · Nanyang Technological University

LLM Alignment Language Model Decoding Optimal Stopping LLM Reasoning

Jun 19, 2026cs.AI

AI Alignment From Social Choice Perspectives

Alignment from human feedback uses human judgments about model outputs to steer the behavior of language models after pretraining. When those judgments reflect conflicting views of desirable behavior, the learned objective becomes an aggregate determination of what the model should prefer. We survey recent work that has studied this aggregation problem through the lens of social choice theory. We illustrate how the social choice perspective helps identify failure modes in the feedback aggregation layer and reveals a broader design space for handling disagreement in explicit and principled ways.

Daniel Halpern, Evi Micha, Ariel D. Procaccia +3

Google Research · University of Southern California · Harvard University

LLM Alignment AI Alignment Multi-Objective Language Model Alignment RL from Human Feedback

Jun 19, 2026cs.CL

Towards Pedagogically Aligned LLM Tutors for Math Mistake Remediation

Large language models have strong potential for use in intelligent tutoring systems, but they often fail to follow effective pedagogical strategies, such as guiding students without revealing final answers. We study the application of a two-stage alignment pipeline for math mistake remediation, combining supervised fine-tuning on tutoring dialogs with Direct Preference Optimization on synthetic preference pairs. We construct a dataset that integrates existing tutoring corpora with synthetic data generated along pedagogical dimensions, such as scaffolding and factuality, and study different input configurations that incorporate solution correctness and gold answers. Experiments show that this approach improves both factual accuracy and pedagogical quality over base models and existing tutoring models. Human evaluation further indicates that our best model is competitive with a strong proprietary baseline, while providing additional benefits in terms of openness, transparency, and reproducibility. Our results highlight the effectiveness of preference-based pedagogical alignment, while also revealing challenges in reliably evaluating tutoring quality.

Kseniia Petukhova, Tien Dat Nguyen, Ekaterina Kochmar

Mohamed bin Zayed University of Artificial Intelligence

LLM Alignment Intelligent Tutoring Systems Generative AI in Education Instruction Tuning

Jun 18, 2026cs.AI

What Shapes Emergent Misalignment? Insights from Training Dynamics, Model Priors, and Data

Emergent misalignment (EM) is a phenomenon in which models generalize with narrow fine-tuning, leading to broad (yet uneven) misalignment across evaluation questions. We study EM and its variability directly through the components of fine-tuning: training dynamics, model priors, and data. (1) We first explored how in-domain training loss relates to out-of-domain alignment scores across datasets and model families. Then, we tried to induce potential alternative local minima through different learning schedules for one narrow fine-tuning, but did not find strong runs with better broad alignment scores conditioned on similar or lower training loss. (2) We found that although the mean and standard deviations of the misaligned model scores are usually statistically different from those of the pre-trained model, there are some potential signals on overall positive correlation. The evaluation prompt-only activations from both the pre-trained and the original instruct models (prior to narrow fine-tuning) could predict fine-grained alignment scores after narrow fine-tuning. (3) Finally, we compared activation deltas before and after narrow fine-tuning and found moderate-to-high subspace overlap and similarity between the resulting activation shifts for training and evaluation prompts. Subspace overlaps between training and evaluation prompt activations correlate with their shifts' similarities when measuring with the last prompt-token activations. The train-evaluation data prompt overlap is controlled against overlap computed from random vectors and evaluation prompts activations.

Yuchen Zhang, Anietta Weckauff, Diego Garcia-Olano +1

Max Planck Institute for Intelligent Systems, ELLIS Institute Tübingen, Tübingen AI Center, Tübingen, Germany · Meta.

LLM Evaluation LLM Alignment LLM Interpretability LLM Fine-Tuning Emergent Misalignment in Language Models Emergent Misalignment

Jun 18, 2026cs.CL

Your Mouse and Eyes Secretly Leak Your Preference: LLM Alignment using Implicit Feedback from Users

To align a Large Language Model (LLM), most existing methods collect explicit human feedback and train a reward model to predict the human preference based on the response text. These existing methods have two key limitations. First, the users rarely provide explicit feedback for LLM responses, which makes the high-quality preference annotation expensive to collect. Second, the methods do not leverage implicit human feedback, which has proven vital to the economic moats of Internet giants. To quantify the value of implicit feedback, we build a new dataset called IFLLM, which collects 1336 multi-turn questions from the 59 Mechanical Turk workers, their mouse trajectories, and eye gazing points to the LLMs' responses from their webcams. IFLLM shows that the users have very diverse types of gazing behavior and mouse trajectories. Our reward model based on the implicit user feedback boosts the accuracy of the text-based reward model from 55% to 64% and nearly triples the relative response quality improvements after applying the DPO to eight LLMs, demonstrating the value of implicit feedback in the wild. Our data collection website, dataset, and codes can be found at https://github.com/themehulpatwari/llm-implicit-feedback/.

Haw-Shiuan Chang, Jeffrey Gomez, Mehul Patwari +2

University of Massachusetts, Amherst, USA · York University, Canada

Reward Modeling LLM Alignment

Jun 18, 2026cs.HC

Editorial Alignment: A Participatory Approach to Engaging Editorial Expertise in LLM-mediated Knowledge Dissemination

The emergence of LLM-driven information services is reshaping the conditions under which public knowledge institutions operate, threatening to absorb the editorial function these institutions exist to exercise. While LLMs offer powerful new affordances for knowledge dissemination, editorial authority is challenged by pretrained LLMs that arrive already aligned with the values and dissemination strategies of their commercial developers. This paper investigates editor participation in re-aligning LLM interfaces to editorial standards through design workshops, in a case study where we design and implement an LLM-enabled encyclopedia interface with a Nordic public knowledge institution. We introduce editorial alignment as a design practice within Participatory AI, framing AI alignment as a design process and positioning the editorial standard as a design artefact that translates editorial practice and values into alignment objectives for technical implementation. Last, we discuss how editorial alignment can create space for ongoing participation and give editors agency in LLM-mediated knowledge dissemination.

Simon Aagaard Enni, Malthe Stavning Erslev, Karl-Emil Kjær Bilstrup +1

Aarhus University Aarhus, Denmark · University of Copenhagen Copenhagen, Denmark

Jun 17, 2026cs.AI

Which Pairs to Compare for LLM Post-Training?

Preference-based post-training has become a central paradigm for aligning language models. A common data-collection strategy is to generate a small set of completions for each prompt and label the resulting comparison pairs. However, human preference labels are often much more expensive than generating additional completions, suggesting a different use of the same labeling budget: generate a larger pool of completions, but label only the most informative comparison pairs. This paper studies which pairs should be compared in preference-based post-training. We formulate comparison curation as a sampling-design problem and evaluate designs by the quality of the final policy under the preference-based post-training objective. We instantiate this framework for Direct Preference Optimization (DPO), analyzing how the choice of labeled pairs propagates through DPO training to downstream policy performance. Our main results provide matching upper and lower bounds on the post-training optimality gap of the DPO-trained policy. The bounds show that comparison selection affects downstream performance through a single design-dependent information matrix, which links label allocation to parameter estimation error and policy suboptimality. This yields an explicit optimization criterion for budgeted comparison curation and motivates practical sampling designs for selecting informative pairs from large generated completion pools. Experiments on synthetic settings and language-model post-training benchmarks show that the proposed designs consistently improve sample efficiency over common comparison-selection heuristics.

Jiangze Han, Vineet Goyal, Will Ma

Columbia University

Pairwise Preference Learning Pairwise Comparison LLM Alignment Pairwise Preference Evaluation Direct Preference Optimization Data Selection+1

Jun 17, 2026cs.LG

Tracking Representation Dynamics in Large Language Models with Persistent Homology

Large language models are commonly aligned through supervised fine-tuning, yet little is known about how their internal representations evolve during this process. We study alignment dynamics using persistent homology by tracking the topology of activation spaces throughout fine-tuning. Across four transformer language models ranging from 1B to 7B parameters and three alignment objectives corresponding to helpful, harmless, and mixed training data, we find that the majority of topological reorganization occurs during the earliest stages of training. A dense checkpoint analysis reveals a transient peak in topological activity followed by rapid stabilization. We further show that different alignment objectives induce distinguishable topological trajectories, while instruction-tuned and pretrained models exhibit qualitatively different patterns of evolution. Our results suggest that persistent homology provides a complementary perspective on alignment, revealing representation-level changes that are not apparent from behavioral metrics alone.

Naman Malhotra, Jay Ambadkar, Abhinav Gupta +4

Department of Mathematics, Imperial College London, UK · Department of Computing, Imperial College London, UK

LLM Alignment LLM Fine-Tuning Representation Geometry in Language Models

Jun 17, 2026cs.AI

Emergent Alignment

Can Large Language Models (LLMs) discern when their own outputs are misaligned with human ethics? And can they self-correct? We endow an LLM with a conscience step that reviews its own reasoning and outputs, and we extend the training loss with an alignment component using Direct Preference Optimization (DPO) to steer the model away from non-ethical outputs. The result is an online technique to align models in a wide range of applications: training, fine-tuning, adversarial prompting, and zero-shot learning. It does not require a weaker or stronger judge, relying instead on a frozen copy of itself. In previous work, the Emergent Misalignment scenario showed a range of emergent unethical behaviors from fine-tuning the model to hack code. Instead, we empirically show how to achieve Emergent Alignment: a single high-level introspective question steers training toward an ethical model under the same code hacking scenario.

Martin Kolář

CIIRC, Czech Technical University in Prague, Prague, Czech Republic.

LLM Self-Correction LLM Alignment Moral Reasoning in Language Models

Jun 17, 2026cs.AI

Beyond Safe Data: Pretraining-Stage Alignment with Regular Safety Reflection

To achieve deeper safety alignment for large language models (LLMs), recent efforts have studied how to push safety interventions earlier into the pretraining stage, primarily by filtering unsafe data or rewriting it into safer forms. We argue that pretraining-stage alignment should go beyond making the data safe: LLMs may compose seemingly benign knowledge and capabilities into unsafe behaviors. To this end, we propose Safety Reflection Pretraining, a pretraining-stage alignment method which regularly inserts short safety reflections into pretraining corpora to integrate self-monitoring directly into language modeling, establishing a foundational capability that is subsequently reinforced by compatible post-training. Our experiments with 1.7B models pretrained on FineWeb-Edu show that Safety Reflection Pretraining improves safety classification accuracy and substantially reduces the success rates of inference-stage and finetuning attacks. Complementary to our real-world experiments, we also introduce a fully controlled synthetic environment, MedSafetyWorld, with a clear definition of safety and a reasoning structure under which models can easily generalize unsafe behaviors from safe data. Ablations in MedSafetyWorld further demonstrate a clear advantage of Safety Reflection Pretraining in preventing models from acting on unsafe behaviors generalized from safe data, compared with data filtering and rewriting. Taken together, our findings suggest that pretraining alignment should not only make the training data safe, but also shape the behaviors that models are likely to acquire from safe data.

Jinhan Li, Kexian Tang, Yihan Xu +2

Institute for Interdisciplinary Information Sciences, Tsinghua University

Language Model Pretraining LLM Alignment LLM Safety Alignment Adversarial Robustness of Language Models