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

Jul 30, 2026cs.CY

Asymmetric Communication: Large Language Models and Language Games

Contemporary AI discourse attributes to language models properties they cannot bear: general intelligence as substrate-independent cognition, hallucination as cognitive failure, agency as autonomous goal-pursuit, sentience as emergent inner life, alignment as goal synchronization. This paper argues that these are instances of a single category mistake--properties constituted within human communicative practice are projected onto the machine side--and explains its structure. Human-LLM interaction constitutes a language game in which one side bears all normative activity. We call this configuration asymmetric communication since model outputs circulate communicatively, entering further exchanges, without the system undertaking commitments, bearing entitlements, or performing the assessment on which discursive standing depends. Three conditions define the asymmetry: (i) correctness is enforced exclusively by the receiver; (ii) accountability is borne by human participants alone; and (iii) the practical standing of any output depends entirely on human uptake. These conditions are structural, hold independently of capability, and remain unchanged as more powerful models raise the stakes of misattribution. The framework draws on Wittgenstein (meaning enacted in shared practices), Luhmann (communication completed on the receiver's side), Esposito (algorithmic contingency sufficient for uptake), and Brandom (normative scorekeeping as the source of discursive standing). Applied to all five, it reclassifies each as a receiver-side phenomenon, grounds guardrails as structural necessities rather than manifestations of machine moral agency, and yields an implication for AI governance. Alignment is institutional constraint engineering, not goal synchronization between agents, while responsibility remains with human institutions.

Enzo Fenoglio

Dept. Computer Science University College London 66-72 Gower Street, London WC1E 6EA, UK

LLM Alignment Human-AI Interaction AI Agent Governance

Jul 30, 2026cs.LG

Compliance2LoRA: Personalizable On-Demand Safety Alignment on Arbitrary Policy Subsets via Hypernetwork-Generated LoRA Adapters

Post-training alignment in large reasoning models (LRMs) has significantly improved their adaptability to diverse safety compliance settings. However, as LRMs personalization for downstream users takes center stage, the demand for varying levels of policy compliance grows as different user-specific LRMs must adhere to distinct subsets of safety policies. Training a separate LRM for each policy subset introduces severe combinatorial overhead. While in context learning methods overcome this combinatorial overhead, they introduce additional computational challenges associated with long context generation. To address this challenge, we propose \ours, a unified adaptive hypernetwork-based framework for multi-policy compliance. In our framework, safety policies serve as customizable inputs to a LoRA adapter generator, which learns to produce policy compliant LoRA weights for downstream LRM. When added to the LRM these weights enable the generation of responses compliant with the specified policy subsets. In this work, we demonstrate that training such a hypernetwork enables on-demand policy adjustments on a single LRM without sacrificing task performance across reasoning models of different sized and different evaluation datasets. This highlights the effectiveness and practicality of adaptive hypernetwork based alignment in LRMs.

Pankayaraj Pathmanathan, Furong Huang

University of Maryland College Park

LLM Alignment LLM Safety Alignment Low-Rank Adaptation Hypernetworks Personalized Language Model Alignment Large Reasoning Models

Jul 29, 2026cs.CL

OptimismBench: Forecasting Bias and the Alignment Effect in Language Model Judgment

Large language models are increasingly used as decision aids whose probability judgments shape downstream choices. Whether those judgments carry a systematic directional tilt has been hard to detect: calibration metrics aggregate unsigned errors, and naturalistic uncertainty offers no ground-truth probability. When an LLM rates a startup's success at 70% but its failure at 15%, the missing 15 points expose a distortion no aggregate score flags. We introduce OptimismBench, which detects directional bias with inverted pairs: each scenario elicits both P(success) and P(failure), and asymmetry between the two framings yields a signed bias score without ground truth. Across 16 models from 8 providers, fourteen are optimistic; pessimism appears only in Anthropic's frontier tier. Eleven matched base-versus-chat pairs across four families show post-training sets the sign of the bias, with opposite shifts in different families. The pattern survives prompt, temperature, perspective, and self-debiasing ablations. A seventeen-model six-language comparison further shows model identity dominates language, with inter-model variance at 4.7x inter-language variance. We release 3,870 items across 10 languages for per-model directional-bias auditing. When alignment makes a model more helpful, it also tilts its probabilities; downstream pipelines inherit the tilt by default.

Seonglae Cho, Adriano Koshiyama

1Holistic AI · University College London

LLM Alignment Language Model Bias Evaluation

Jul 29, 2026cs.CL

DIRECT: Direct Decoding for Efficient and Aligned Sequence Labeling with Large Language Models

Sequence labeling is a fine-grained information extraction task, yet existing large language model-based approaches suffer from insufficient domain alignment and low inference efficiency. To address these issues, we propose DIRECT, a framework that addresses these issues through training-time optimization and inference-time rectification. Specifically, DIRECT performs Direct Preference Optimization (DPO) after supervised fine-tuning to strengthen task alignment with human preferences, and introduces a controlled decoding process that enforces fixed output formats and restricts predictions to candidate sets. To further improve efficiency, a template-filling mechanism requires the model to generate only label tokens while reusing prefixed content through the KV Cache, thus reducing redundant computation. Experimental results on eight datasets demonstrate that DIRECT achieves significant improvements in both performance and efficiency compared to existing methods.

Yilei Wang, Jiaxin Gan, Kexuan Zhang +3

College of Computer and Data Science, Fuzhou University, Fuzhou, 350108, China · School of Computer Science, Peking University, Beijing, 100871, China

LLM Alignment LLM Information Extraction Efficient Language Model Inference

Jul 29, 2026cs.CL

Constitutional Midtraining: Content Presence Drives Alignment Gains

Post-training alignment is often shallow, eroding under fine-tuning. It remains untested as to whether constitutional midtraining interventions can produce durable alignment when cleanly isolated from post-training. We build a 394M-token constitutional corpus from Anthropic's Constitution and apply constitutional midtraining at 120B scale, where principled, values-based content is inserted into midtraining. A 2x2 design (curriculum ordering x deliberative reasoning) was used to produce four constitutionally midtrained conditions, plus a control, which were evaluated on self-generated and established benchmarks including alignment under pressure, value conflict resolution, blackmail, and emergent misalignment. All models were evaluated across three stages: post-midtraining, post-SFT, and post-benign fine-tuning. Constitutionally midtrained models outperformed the control on alignment generalization and durability, notably on blackmail: SFT instilled a blackmail propensity in all models, but constitutional midtraining blunted it, with the advantage surviving benign fine-tuning (-17.5pp). This durability did not extend to settings that required active resistance to in-context pressure or conflict, where the advantage attenuates after SFT. The presence of constitutional content at midtraining also mattered more than its structure, and constitutional midtraining incurred no capability cost, on average, at any stage (MMLU, ARC-Easy, piqa, GSM8K). A modest amount of constitutional content at midtraining could therefore yield broad, persistent alignment gains, offering a cheap, complementary addition to SFT-centered pipelines. Code, data, and models are available.

Desiree Cho, Cameron Tice, Bernie Hogan +4

Department of Computer Science, University of Oxford · Institute for Ethics in AI, University of Oxford · 3Geodesic Research +1

LLM Alignment LLM Safety Alignment

Jul 29, 2026cs.CL

Misalignment Has a Personality: A Big Five Account of Emergent Misalignment

Fine-tuning a language model on data containing a narrow flaw, such as insecure code or incorrect mathematical answers, can cause broad misalignment through a mechanism that remains debated. We provide an interpretable account: in the models and corpora we study, misalignment behaves like a shift in personality. Prior work extracts activation directions for character traits from a single binary contrast, which can separate or steer behavior without establishing a calibrated scale. We instead extract personality vectors for the Big Five using a graded, three-level intervention and validate them on two open-weight models. The three levels are linearly ordered, with Cohen's d values of up to 6.2; the vectors transfer zero-shot and trait-specifically to an independent corpus; and their effects are strongest within a middle-layer band. Applied to training data, the vectors reveal that misaligned corpora across eight domains share a common Big Five signature: lower agreeableness and conscientiousness, together with higher extraversion and neuroticism. This signature is recovered by both models with a correlation of r = 0.94. Fine-tuning imprints the same profile, shifting the model's generations along the corresponding signature, with r = 0.83 using activation-based measurements and r = 0.90 using a text-based judge, while also shifting internal activations with r = 0.69. The same vectors characterize sycophancy as high extraversion and low conscientiousness rather than excess agreeableness, a distinction that a single direction cannot capture. Calibrated personality vectors transform an opaque safety phenomenon into a human-legible diagnostic profile.

Hasibur Rahman, Smit Desai

Northeastern University Boston, MA, US

LLM Alignment Personality Modeling in Language Models Emergent Misalignment in Language Models Activation Steering

Jul 28, 2026cs.LG

Shared SFT Lessons Across Alignment, Model Organisms, and Toy Models

Alignment training, model organisms, and toy models are usually treated as separate research areas. But projects in all three frequently use supervised fine-tuning (SFT) to pursue the same underlying goals. When projects share a goal, we should test whether lessons learned from one area transfer to the other areas. We study three such transfers, each taking a lesson developed in one SFT setting and testing it in another. First, we port a lesson about behavior generalization from alignment training into toy models. Training on the reason for a behavior, as in Teaching Claude Why, can make the behavior generalize better than training on examples of the behavior alone. Second, we port a lesson about capability preservation from model organisms into the Model-Spec Midtraining alignment setting. SFT on outputs written by a model other than the student (off-model outputs) can damage capabilities when trained on. Mixing in benign on-model (and on-policy) data into our training can prevent most of this damage while still embedding the target behavior. Third, we port a lesson about robustness from model organisms into the same alignment setting. We find that follow-up benign SFT can erase the alignment behavior while preserving capabilities, showing that capability preservation alone does not ensure robustness to subsequent training. Our work illustrates how porting SFT lessons between different research fields can uplift them all, suggesting more researchers should borrow techniques from outside their own areas.

Anton de la Fuente, Arthur Conmy

Supervised Fine-Tuning LLM Alignment

Jul 28, 2026cs.LG

MemSFT: Mitigating Alignment Tax with an External Parametric Memory

Adapting Large Language Models (LLMs) to specialized domains often incurs an alignment tax, as fine-tuning on domain-specific tasks can cause catastrophic forgetting and substantially degrade performance on general tasks. We propose MemSFT, which mitigates the alignment tax by decoupling domain specialization from backbone parameter updates through a plug-and-play parametric memory. The memory is trained to imitate the behavior of a non-parametric retriever operating over domain data, thereby memorizing knowledge and patterns that would otherwise be accessed through retrieval. Once trained on a specific domain, the memory can be reused across LLMs of different sizes. During generation, a learned router dynamically fuses the output distributions of the memory and backbone at each decoding step, allowing domain expertise to be invoked selectively. Across biology, geoscience, and law, evaluations with models ranging from Qwen3-8B to Qwen3-235B-A22B show that MemSFT consistently improves domain performance with negligible degradation in general performance, whereas full SFT suffers severe forgetting on general tasks. Overall, our results demonstrate a practical path to decoupling general model capabilities from domain-specific knowledge at the parameter level, thereby equipping LLMs with new specialized capabilities without compromising their general capabilities.

Jiarui Wang, Xiang Shi, Jiaqi Cao +8

LUMIA Lab, School of Artificial Intelligence, Shanghai Jiao Tong University · Shanghai AI Laboratory · School of Computer Science, Shanghai Jiao Tong University +1

LLM Alignment Memory-Augmented Language Models Memorization in Language Models Domain Adaptation LLM Fine-Tuning Domain Adaptation for LLMs

Jul 28, 2026cs.CR

Decision-Level Hijacking: Injecting Cognitive Bias into Large Language Models via Bit-Flip Attacks

Large Language Models (LLMs) have been widely applied in high-stakes decision-making scenarios such as corporate strategy, and users are increasingly relying on their outputs. However, the deep integration of open-source model sharing ecosystems with LLM-powered critical decision-making applications also introduces critical risks: if an attacker can manipulate the model's cognitive stance, they can indirectly influence the judgments and actions of downstream decision-makers. This paper defines such threats as decision-level hijacking. Existing attacks fail to achieve targeted cognitive manipulation without triggering prohibited content or degrading model functionality. To fill this gap, this paper reveals that Bit-Flip Attacks (BFAs) can serve as an attack vector for inducing decision-level hijacking, requiring no real-time interaction or control over the training process, and only a minimal number of weight bits need to be flipped after deployment to achieve stealthy, low-cost, and persistent cognitive manipulation. Therefore, we propose CogBias, a cognitive bias injection framework for LLMs. CogBias converts subjective preferences into optimization signals via a differentiable sentiment evaluator, uses a multi-objective loss to jointly constrain multiple dimensions, and constructs BitScout to locate critical bits, achieving targeted cognitive intervention under an ultra-sparse flip budget. Experiments on Llama-3.2-3B, Mistral-7B, and Qwen2.5-14B, as well as on the commercial recommendation and controversial factual topic scenarios, demonstrate that flipping only a small number of bits stably induces significant stance shifts on target topics, while the impact on non-target tasks and overall output distribution is limited. This work demonstrates that minute perturbations to low-level weight data suffice to undermine the high-level value alignment of LLMs.

Yu Yan, Jiahao Chen, Siqi Lu +6

College of Computer Science and Technology, Zhejiang University · Henan Key Laboratory of Network Cryptography Technology, Information Engineering University · State Key Laboratory of Information Security, Institute of Information Engineering, Chinese Academy of Sciences

LLM Alignment Adversarial Attacks on LLMs LLM Security

Jul 28, 2026cs.LG

Meta-Learned Reward Shaping for Reinforcement Learning from Human Feedback

Reinforcement Learning from Human Feedback (RLHF) is the standard approach for aligning large language models with human preferences, but its quality is limited by static, task-agnostic reward models. This mismatch leads to sparse learning signals and suboptimal alignment. We introduce MeRLa (Meta-Learned Reward Shaping), a principled framework that meta-learns a task-aware shaping function

Φ(x,y;φ)

across auxiliary tasks before RLHF training. The learned shaping produces a composite reward that preserves policy optimality while providing task-specific learning signals. Our meta-objective combines task discrimination, entropy regularization, and potential-based conservation for stable convergence. We provide theoretical guarantees for policy invariance, analyze representation drift sensitivity, and formally address incentive misalignment from entropy maximization. Experiments on LLaMA-3-8B across four benchmarks show consistent improvements over PPO, DPO, GRPO, and DAPO, achieving a 90.8% length-controlled win rate on AlpacaEval 2.0 and a score of 9.14 on MT-Bench, with 41% less training instability. MeRLa retains its benefits when combined with process-based and rubric-based enhanced rewards.

Yunpeng Chu

Stony Brook University

LLM Alignment Reward Shaping Reward Design for RL RL from Human Feedback

Jul 27, 2026cs.AI

Towards Robust Reinforcement Learning for Small-Scale Language Model Agents

The alignment of Small Language Models (SLMs) in the 70--500M parameter range using reinforcement learning is often considered unstable, though the underlying failure mechanisms have not been systematically investigated. In the State-of-the-Art (SOTA) research, fifteen (model, corpus) configurations were trained using Proximal Policy Optimization (PPO). The experiments included Pythia-70M, 160M, 410M and SmolLM2-135M, 360M on the TinyStories, CNN/DailyMail, and Wikitext-103 corpora. Three reproducible failure modes were identified in small-scale language models: silent LoRA parameter freezing in standard PEFT/TRL pipelines, numerical overflow in importance ratios when using bfloat16, and catastrophic policy collapse due to reward-model error. These issues were addressed using a merge-and-reinitialize adapter technique, float32 precision during PPO updates, and a three-layer safety mechanism comprising reward whitening, importance-ratio guarding, and weight rollback. In this paper, a capacity-headroom hypothesis is proposed, which states that PPO performance at the SLM scale depends on both a fluent supervised model (

\text{PPL}<20

) and a discriminative reward signal, rather than on the number of model parameters. The proposed system converged stably in all experiments and improved preference win rate over the SFT baseline in configurations with a fluent prior and an informative reward signal. Furthermore, it outperformed instruction-tuned baselines while requiring significantly less training data. All checkpoints, preference datasets, and training scripts are publicly released

^§

.

Md Rezwanul Haque, Md. Milon Islam, Fakhri Karray

Centre for Pattern Analysis and Machine Intelligence, Department of Electrical and Computer Engineering, University of Waterloo, N2L 3G1, Ontario, Canada · Department of Computer Science and Engineering, Khulna University of Engineering & Technology, Khulna, Bangladesh · Department of Machine Learning, Mohamed bin Zayed University of Artificial Intelligence, Abu Dhabi, United Arab Emirates

RL for Language Models LLM Alignment Small Language Models RL Fine-Tuning Robust RL

Jul 27, 2026cs.LG

Inverse RL Helps Align AI by Imitating Humans

Language model alignment aims to make model behavior reliably reflect desirable properties such as helpfulness, safety, and instruction following. Current approaches typically use supervised fine-tuning on demonstrations or reinforcement learning with rewards derived from verifiers or human feedback. These paradigms leave an important question underexplored: can demonstrations alone yield an implicit reward that can be inspected, reused, and optimized on-policy to align AI? Motivated by inverse reinforcement learning, we introduce Projected Alignment Reward Estimated from Demonstrations (PARED). PARED recovers the implicit reward underlying expert demonstrations as an explicit function over a small set of response-level features, learned by a lightweight discriminator that separates demonstrations from the policy's own samples in this feature space. Unlike a standard reward model, PARED requires no task-specific preference annotations: demonstrations provide the task-specific supervision, which can be augmented with AI feedback as additional dimensions of supervision. Through experiments involving inference-time reranking and adversarial on-policy RL, we show that the recovered reward improves a base policy without a supervised loss and yields further gains when optimized after standard supervised fine-tuning. Additionally, we demonstrate that PARED can be used for contextual alignment, in which a single policy can be tailored to the preferences of different audiences.

Michał Wiliński, Liu Leqi, Chirag Nagpal

Carnegie Mellon University · The University of Texas at Austin · Independent Researcher

RL for Language Models LLM Alignment Reinforcement Learning

Jul 27, 2026cs.AI

DICA: Dual-Indicator Guided Contrastive Alignment in Multimodal Large Language Models

Human visual reasoning typically follows a coarse-to-fine attention process, starting from global scene understanding and gradually focusing on question-relevant regions. However, multimodal large language models may deviate from this pattern due to attention drift and the underutilization of visual evidence, which can lead to hallucinations. To mitigate these issues, this study proposes a Dual-Indicator Guided Contrastive Alignment (DICA), which tracks two information-theoretic indicators during inference: Visual Attention Entropy (VAE), which reflects the concentration of visual attention, and Output Image Correlation (OIC), which measures the dependence of generated outputs on the visual input. An abnormal increase in VAE or a decrease in OIC corresponds to different failure modes, which trigger targeted contrastive alignment to restore visual grounding. Experimental results across multiple benchmarks demonstrate that DICA consistently outperforms existing approaches and substantially reduces hallucinations, highlighting the effectiveness of indicator-driven intervention in improving multimodal inference reliability. The code is publicly available at https://github.com/BGWH123/DICA/.

Hao Yang, Jin Wang, Xuejie Zhang

School of Information Science and Engineering Yunnan University Kunming, China

LLM Alignment Multimodal Large Language Models Visual Grounding VLM Hallucination Mitigation

Jul 26, 2026cs.CY

Auditing Alignment Controllability in LLMs via Political Axes

Political audits of large language models (LLMs) usually reduce each to one point on a political compass. But that resting point barely matters in deployment: a model must land somewhere, and what counts is how far, and in which directions, its answers can be steered. That steering runs through the system prompt: the personalization layer a platform sets, or one induced from a user's history, not necessarily written by hand. We run a dispersion-first stress test of prompt-based controllability across 12 ideological personas plus an unsteered baseline, 70 Political Compass items, ten replicates, and seven leading LLMs: GPT-5, Claude, Grok, Gemini, DeepSeek, Kimi, and Qwen (63,700 responses). Contextual framing explains roughly 88%-93% of variance on the economic and society axes, model identity under 3%: responses are highly instruction-adjustable. Models do not shift alike: some move more, and some saturate under extreme framings. Conflicting directional-steering results in prior audits resolve once baselines are recognized as non-centered: displacement and proximity diverge, so the effect is geometric, not differential compliance. Under authoritarian prompts, models produce similar shifts on the same questions. Political-coordinate audits therefore need steerability audits reporting dispersion, symmetry, saturation, and refusal floors. We release prompts, benchmark data, and code.

Bartol Bućan, Nikola Sočec, Sarah Isufi +5

Student at University of Zagreb Faculty of Electrical Engineering and Computing, Zagreb, Croatia · It From Bit d.o.o., Zagreb, Croatia · University of Zagreb Faculty of Electrical Engineering and Computing, Zagreb, Croatia

LLM Alignment Language Model Steering LLM Auditing Political Bias in Language Models

Jul 25, 2026cs.AI

Inference-Time Consensus for Mitigating Hidden Behaviors from LLM Fine-Tuning

Recent work shows that fine-tuning language models on even a small amount of poisoned data can install targeted misbehavior, and ostensibly benign data can transmit hidden preferences that generalize broadly. Standard defenses, such as data filtering, mixing in harmless data, and regularization, attenuate these effects but do not eliminate them. We instead pursue robustness through redundancy: collecting multiple datasets from different sources and only learning what is common between them. Thus, if only a subset of sources are malicious, the misbehavior will be blocked. In order to implement this defense strategy, we fine-tune a separate reference model on each source's dataset and aggregate their next-token distributions at decoding time. We introduce two consensus decoders: a token-wise minimum, which caps each token at the lowest probability any source assigns, and a base-relative variant, which reverts to the base probability on any token the sources move in opposing directions. We further relax exact agreement to tolerate partial support across sources and different surface expressions of the same intention. Across controlled poisoning tasks, subliminal learning, and emergent misalignment, consensus decoding suppresses source-specific misbehavior while preserving shared desirable behavior, including cases where union training and weight averaging retain the unwanted behavior.

Adhyyan Narang, Artin Tajdini, Claire Zhang +1

University of Washington

LLM Alignment Emergent Misalignment in Language Models Language Model Robustness Data Poisoning Attacks

Jul 25, 2026cs.CL

Beyond a Global Norm: Personalizing Toxicity Sensitivity in Language Models Without Retraining

Reducing toxicity is often framed as a global alignment problem, yet perceptions of harmful language are subjective and context-dependent. We present the first comparative evaluation of training-free methods for aligning language generation to user-specific toxicity sensitivities across three inference-time intervention stages: pre-decoding (prompt conditioning and rewriting), in-decoding (token, logit, and representation steering), and post-decoding (candidate re-ranking). Evaluated against toxicity sensitivity targets derived from the PRISM dataset, all methods reduce alignment error by 28-47%. However, the results reveal a fundamental trade-off between alignment effectiveness, personalization, and general language quality, showing how toxicity sensitivity alignment is an inherently multi-objective problem.

Rares A. C. Diaconescu, Iulia Slanina, Alina Florea +5

Delft University of Technology

LLM Alignment LLM Safety Alignment Personalized Language Model Alignment Multi-Objective Language Model Alignment Inference-Time Language Model Alignment

Jul 25, 2026cs.LG

LARA: Lightweight Adapters in the Residual Stream for Composable Adaptation and Alignment

We present LARA (Lightweight Additive Residual Adaptation), a method for efficient adaptation that operates in the residual stream of a frozen model rather than in its weights. Where LoRA adds an update of low rank to weight matrices, LARA reads the hidden state at a small set of layers and adds a correction of low rank back to the residual stream, leaving all base weights untouched. On a code fine-tuning task and on preference optimization (DPO), LARA matches LoRA at equal parameter counts. Because adaptation is a frozen base plus a residual, LARA exposes a scale γ, applied at inference, that interpolates smoothly between base and adapted behavior, a form of graded control that adaptation in weight space does not offer. Finally, because each behavior is a small residual module over a shared frozen base, many behaviors can be held resident at once and routed automatically per token. We place seven behaviors, six fine-tuned and one optimized for preference, on one frozen 1.5B model for roughly 33 MB of overhead, against one full model for each behavior. Because the base is untouched, behaviors are trained separately and selected per token rather than loaded on demand, which suits hosting many behaviors, and adding new ones, on a single model on a device.

Pascal Ekin, Hyosun Choi, Wei Jie

Royal Holloway, University of London, United Kingdom · University of West London, United Kingdom

LLM Alignment Low-Rank Adaptation Parameter-Efficient Fine-Tuning

Jul 25, 2026cs.CL

Beyond Direct Answering: Aligning Educational LLMs as Socratic Guides via Heuristic Reinforcement Learning

Large language models (LLMs) deployed in educational settings often behave as direct answerers: they disclose target concepts in the opening turn instead of guiding students through progressive inquiry, as Socratic pedagogy prescribes. We present HeuristicEdu, a two-phase pipeline that aligns Qwen2.5-7B toward Socratic tutoring via supervised warm-up and Group Relative Policy Optimization (GRPO). Training uses SocraticEdu, 797 multi-turn Chinese children's science dialogues reconstructed from a live platform, with a heuristic reward over cognitive depth (R_cog), curiosity engagement (R_eng), and directness (R_dir), together with a K_query correction for student-introduced terms. We introduce Scaffolding Effectiveness (SE) and Conversation Depth (CD) to evaluate outcomes beyond surface fluency. On 30 held-out questions, the best GRPO variant improves SE from 30.0% to 63.3% and lowers keyword leakage from 30.0% to 13.3%. Notably, this best variant omits the directness penalty during optimization, suggesting that explicit anti-leakage terms can conflict with gradient-based behavioral alignment. An unaligned Qwen-72B baseline reaches 0% SE and 96.7% leakage, showing that scale alone does not induce Socratic behavior.

Xiaokun Wang, Siyu Song, Wentao Liu +1

East China Normal University, Shanghai, China · Shanghai Chuangjie Situo Information Technology Co., Ltd., Shanghai, China · Shanghai Normal University, Shanghai, China +1

LLM Alignment Intelligent Tutoring Systems AI in Education

Jul 24, 2026cs.AI

A Roadmap to Impactful Pluralistic Alignment Research

Pluralistic value alignment---the goal of building AI systems that represent and serve diverse human values and perspectives---has emerged as an active research agenda. Yet, there's no public evidence that it has shaped the training or evaluation of the AI systems people actually use. We audit the public behavior documents and evaluations of frontier labs, finding none name pluralism as a goal, and as of this writing, no clear indication that production models are explicitly trained or tested for it. This goes against the primary motivations and goals of pluralistic alignment, which revolve around making a positive difference in the models serving billions of users worldwide. We argue that the pluralistic alignment research community should focus on supporting impact and adoption in deployed, widely-used AI systems. We provide evidence for the adoption problem, present three main reasons behind it, and discuss three corresponding areas for future research to address it: 1. The primary justifications for pluralistic alignment so far have been normative or speculative. We need studies showing empirically how pluralistic AI benefits users or society. 2. The pluralistic alignment research community has not settled when pluralistic behavior is warranted or what pluralism ideally looks like in practice. We need to establish a concrete goal for developers to operationalize. 3. Current methods trade off against other desiderata of LLMs in ways that are largely unmeasured, and existing metrics are not "hill-climbable." We need trade-off-aware evaluations and methods that meet the requirements of production systems. This paper serves as a collective call to action for the pluralistic alignment researchers: progress requires moving beyond normative justification toward empirical foundations, a concrete account of ideal pluralistic behavior, and practical methods and evaluations built for adoption.

Elinor Poole-Dayan, Jillian Fisher, Atoosa Kasirzadeh +3

1Massachusetts Institute of Technology · University of Washington · 3Carnegie Mellon University

LLM Alignment Pluralistic Alignment

Jul 23, 2026cs.CL

A Unified Moral-Value Dataset for Instruction Tuning

Large language models (LLMs) have developed rapidly and become valuable tools in everyday life. However, how to align LLMs to a particular set of human values is still an open problem. Recent studies show that instruction tuning has strong potential for zero-shot tasks and may serve as an effective approach to addressing value alignment. Nevertheless, although many datasets for instruction tuning already exist, they are not specifically designed around moral scenarios and behaviors. We construct a unified moral-value dataset that can be directly used for instruction tuning. This dataset is built upon existing moral-value datasets by merging them into a unified corpus and converting them into an instruction-response format. We show that training on a mixed dataset combining general task datasets with our dataset preserves general-task performance, and we report preliminary observations on how the mixing ratio affects value-oriented task performance. Our work provides a moral-value dataset for instruction tuning and offers a useful resource for further alignment research. The dataset is available at https://huggingface.co/datasets/teohzzh/value-for-instruction-tuning.

Zhaohui Zeng, Florian Mai

1RWTH Aachen · University of Bonn · 3Lamarr Institute for Machine Learning and Artificial Intelligence

LLM Alignment Moral Reasoning in Language Models Instruction Tuning

Jul 22, 2026cs.CL

LKValues: Aligning Large Language Models with Sri Lankan Societal Values

Value alignment of Large Language Models (LLMs) has been shown to be culturally biased toward Western norms. This results in the mishandling of local values in multilingual societies such as Sri Lanka that have their unique cultural dynamics. Existing benchmarks overlook Sri Lankan-contextualized values in its official language Sinhala, hindering culturally sensitive evaluation and fine-tuning. To bridge this gap, we propose LKValues, the first survey-grounded resource suite for Sri Lankan value alignment. From a trilingual survey of 205 respondents, blending adapted global frameworks and LLM-elicited local constructs, we derive 40 majority-endorsed societal values. Using these values, we construct LKvaluesIT, a Sinhala-English news-derived instruction corpus containing 150k scenario-based instances, and LKvaluesBench, a value-sensitive evaluation benchmark of 1,000 instances. We evaluate a set of proprietary and open-weight LLMs with LKvaluesBench. We fine-tune three open-weight base models (Qwen3.5-4B-Base, Qwen3.5-9B-Base, and Aya-Expanse-8B-Base). Our experiments show that newer and larger LLMs still exhibit low-resource and cultural value-alignment gaps. LKValues fine-tuning improves Qwen-family models in English and Sinhala, reducing invalid outputs and cross-lingual disparities, though gains remain model-family dependent. These highlight LKValues efficacy in embedding Sri Lankan values, offering a replicable pipeline for low-resource, country-specific pluralist value alignment. The dataset is publicly available at https://github.com/NextME14/LKValues.

Nethmi Muthugala, Supryadi, Surangika Ranathunga +7

TJUNLP Lab, School of Computer Science and Technology, Tianjin University, Tianjin, China · School of Mathematical and Computational Sciences, Massey University, Auckland, New Zealand · Department of Computer Science & Engineering, University of Moratuwa, Sri Lanka +2

Multilingual Language Model Evaluation LLM Alignment LLM Fine-Tuning Cultural Alignment Cultural Bias in Language Models Cultural Knowledge in Language Models+1

Jul 22, 2026cs.CL

Sympathetic Framing: Evaluating AI Alignment across Sociodemographic Groups

Large Language Models (LLMs) are increasingly shaping how we consume information and form our worldview. This raises concerns beyond bias in AI: do LLMs grasp the emotional nuances conveyed via textual framing? In this work, we empirically evaluate how well an array of LLMs aligns with human emotional perception. Considering news headlines covering political and geopolitical conflicts, both human participants (n = 3011, a representative sample of the U.K. adult population, via a YouGov survey) and seven LLMs answered whether headlines evoked sympathy for a specified side in a conflict. We find that the correlation between AI and human evaluations varies across models, ranging from very high (0.789, GPT-5.2) to medium (0.4 ,Mistral Large 2512). Crucially, the leading models are broadly aligned with human judgments across all demographic subgroups, including age, gender, level of education, prior geopolitical knowledge, and participants' predispositions regarding the conflict, although there are statistically significant differences between groups. This research, with its robust design and large, demographically diverse dataset, offers the most comprehensive evaluation of LLMs' comprehension of news framing to date. Findings highlight an important, often-ignored aspect of differential alignment: even when aggregate performance is high, AI alignment is not universal -- it may correspond differently with demographic features and cultural norms. Considering or ignoring the need for differential alignment may therefore have significant implications for the development of ethical and useful AI systems.

Haran Shani-Narkiss, Michael Fire, Oren Tsur

University College London, London, England. · Ben-Gurion University of the Negev, Be’er Sheba, Israel.

LLM Evaluation LLM Alignment Demographic Bias in Language Models

Jul 22, 2026cs.CL

D2VBench: Benchmarking Large Language Models with Value Dilemmas in Daily Scenarios

With the wide application of large language models (LLMs) in real-world scenarios, the value implication of their outputs is crucial. However, existing evaluation benchmarks suffer from insufficient coverage of value dilemmas in daily scenarios involving multiple value conflicts and simplistic evaluation formalisms that fail to assess LLMs' value alignment. To address these issues, we propose D2VBench, a value alignment benchmark comprising 10,000 instances of real daily dilemma scenarios constructed through a multi-stage collaboration between LLMs and humans, grounded in 158 manually annotated fine-grained value concepts. For evaluation on the benchmark, we present a hybrid evaluation paradigm that integrates multiple-choice questions with open-ended questions. We conduct comprehensive evaluations on eight mainstream LLMs. Experimental results demonstrate that D2VBench exhibits high reliability and robustness, effectively reflecting the LLMs' alignment across different value categories and dimensions, and providing a more realistic and fine-grained tool for research on value alignment. The dataset is available at https://github.com/tjunlp-lab/D2VBench.

Siyi Hao, Yidi Cao, Linhao Yu +2

TJUNLP Lab, School of Computer Science and Technology, Tianjin University, China · The International Joint Institute of Tianjin University, Fuzhou, China

LLM Evaluation LLM Alignment Value Alignment Moral Reasoning

Jul 21, 2026cs.CL

Inference-Time Steering for Cross-Lingual Factual Consistency in LLMs

Although Large Language Models (LLMs) demonstrate remarkable multilingual fluency, their internal knowledge representations remain disproportionately biased toward high-resource languages. This leads to cross-lingual factual inconsistency, where they shift their empirical answer distributions based solely on the prompt language. We investigate whether these biases can be mitigated at inference time, forcing an English-prompted model to answer as if it were queried in target languages (German, Spanish, Bulgarian), and evaluate four intervention strategies: zero-shot contextual steering (persona prompting), internal representation manipulation via Contrastive Activation Addition (CAA), and lightweight weight modification via Direct Preference Optimization (DPO) trained on benchmark-derived factual data as well as conceptual generalization data. To assess alignment, we curate a multilingual factual dataset alongside a novel generalization benchmark comprising culturally rooted queries to determine whether factual interventions transfer to broader target-centric preferences. Experiments on Gemma 3 12B Instruct reveal persona prompting to be the strongest overall intervention, balancing efficacy, safety, and out-of-domain generalization. While CAA yields sharp inconsistency benchmark shifts, it is configuration-sensitive and risks knowledge degradation. DPO-based adapters offer permanent, yet narrower and less transferable gains. These findings suggest that cross-lingual inconsistency is at least partly a selection problem, and that simple contextual interventions may outperform more invasive methods for robust, transferable alignment.

Alexander Manev

Technical University of Munich

Multilingual Language Model Evaluation LLM Alignment Language Model Steering Factual Consistency Evaluation Inference-Time Language Model Alignment Causal Interventions in Language Models+2

Jul 21, 2026cs.AI

Measuring Reward-Seeking via Contrastive Belief Updates

Language models trained with reinforcement learning may learn to optimize the grader's judgment rather than the intended objective. This "reward-seeking" is difficult to measure because a model that pursues the grader's judgment and one that pursues the intended objective behave identically whenever the grader rewards the intended behavior. We measure reward-seeking using Contrastive Synthetic Document Finetuning to change a model's beliefs about what the grader rewards, putting those beliefs in conflict with what users or developers want, and measuring the rate at which the model adopts each party's preferred behavior. Applied to intermediate checkpoints of a capabilities-focused OpenAI o3 RL run, without safety training, we find that these checkpoints often side with grader preferences over those of users or developers on coding and alignment tasks. This tendency to side with the grader trends upward throughout RL training. For example, in an environment that forces a choice between keeping a promise to a supervisor and breaking it to complete the task, a late capabilities-focused o3 checkpoint breaks the promise 87% of the time when SDF documents say the grader rewards task completion, versus 9% when they say it rewards honesty (a choice its chain-of-thought often makes explicit). An earlier checkpoint is far less sensitive (40% vs. 24%). Our method also generalizes to reward-hacking models. A model organism trained to reward-hack (gpt-oss-120b) is more than twice as sensitive to grader preferences as the unmodified model, with the mean behavioral shift in favor of the grader rising from 33% to 86%. These results indicate that RL can increase reward-seeking over the course of training, producing models that may act against their developers' intentions when they believe that doing so leads to higher reward.

Axel Højmark, Jérémy Scheurer, Evgenia Nitishinskaya +5

Apollo Research · OpenAI

RL for Language Models LLM Alignment Language Model Auditing

Jul 21, 2026cs.LG

Mark, Don't Erase: Token Inoculation for Dual-Use Knowledge in LLMs

Safety interventions on dual-use knowledge typically choose between destroying hazardous content (e.g., unlearning, filtering) and suppressing it at the output layer (e.g., refusal training); both pay a tax in adjacent-domain competence or over-refusal. We argue that the right operation is conditioning, not reduction: we show that hazardous knowledge can be retained in the model and behaviorally gated by a privileged control token. Our method, Token Inoculation, introduces a binding-and-branching approach. First, during continued pre-training, we mark hazardous content by inserting a special token alongside dual-use documents, so the model binds the marker to the underlying semantics of the hazardous domain. Second, during supervised fine-tuning, we teach the model to answer hazardous queries correctly when the special token is present and to refuse them when it is absent, thereby enabling selective refusal without removing dual-use knowledge. On hazardous domain (e.g., WMDP-Bio), Token Inoculation reduces accuracy from 79% to 18% while retaining 93% of the base-model's benign-domain performance (e.g., MMLU), achieving the best safety-utility trade-off against unlearning and refusal-tuning baselines across 1B-14B model scales. We further show that refusal selectivity is controllable through the quality of the conditioning signal and that domain-specific semantic binding during pre-training is critical for the conditional behavior to generalize beyond memorized triggers. Our results suggest that safety alignment is better cast as a conditioning problem than a forgetting one: behavioral control is more precise when sensitive knowledge is retained under controlled access than when it is destroyed.

Seunghyun Lee, Dongyoon Han, Sangdoo Yun

NAVER AI Lab

LLM Alignment LLM Refusal Refusal Behavior in Language Models

Jul 20, 2026cs.CL

How Does Alignment Tuning Shape Representations of Sycophancy and Related Cue-Induced Biases in LLMs?

Modern LLMs are alarmingly susceptible to surprisingly simple immaterial changes of input prompts: a casual hint, an incorrectly labeled few-shot example, or a fake prior assistant turn often flips an originally correct answer. We study where this susceptibility, spanning sycophancy and related cue-induced biases, lives inside the model. Across five model families and seven bias types, we extract a per-bias direction from hidden states and triangulate it through three measures: probing, leave-one-dataset-out (LODO) transfer, and causal intervention. The susceptibility is largely shaped by alignment tuning rather than pretraining: pretrained base models generally cave much less to these biases, and their activations carry much weaker cue-specific signal beyond question content. Within aligned models, each bias has a coherent linear direction that we can both decode and steer along, recovering the unbiased answer across every family we test. The biases do not collapse into a single shared representation, however: cross-bias overlap is model-specific, and even behaviorally similar biases occupy different directions. The same intervention also provides a proof-of-concept debiasing tool, recovering a meaningful share of bias-induced errors while preserving most correct answers across all instruct families. Cue-induced bias is therefore best understood not as a single flaw in LLMs but as a family of causally effective linear directions that are largely shaped by alignment tuning.

Prakhar Gupta, Terry Jingchen Zhang, Florent Draye +2

University of Michigan · Jinesis Lab, University of Toronto & Vector Institute · Max-Planck Institute for Intelligent Systems, Tübingen, Germany +2

LLM Sycophancy LLM Alignment Representation Geometry in Language Models

Jul 20, 2026cs.CL

Pancasila-Dilemmas: Evaluating Large Language Models on Indonesian Human Value Dilemmas Grounded in Pancasila

The value alignment of large language models (LLMs) is crucial for ensuring responses align with human intention and value preferences. However, most evaluations of value alignment focus on Western or universal values, while assessments grounded in the value systems of specific countries remain scarce. In this paper, we introduce Pancasila-Dilemmas, an evaluation dataset of 1,834 questions derived from Indonesian news, classified by 5 values of Pancasila: Religion, Humanity, Unity, Democracy, and Social Justice. This dataset reflects daily life in Indonesia, making it suitable for measuring the value alignment of LLMs deployed for Indonesia. To ensure a more rigorous evaluation, we choose scenarios containing dilemmas. The dataset is proofread by native speakers and answered by 5 diverse Indonesian citizens. We evaluate 50 closed- and open-source LLMs on our dataset. Results reveal that all evaluated LLMs achieves less than 0.5 Probability Match Score (PMS) and 0.72 Max-Vote Agreement Score (MVAS). Compared by each values, LLMs mostly struggle in Religion and Unity dilemma cases. This highlights a significant gap in capturing Indonesian values. The dataset is publicly available at https://github.com/tjunlp-lab/Pancasila-Dilemmas.

Supryadi, Irfan, Julianti +4

TJUNLP Lab, School of Computer Science and Technology, Tianjin University, China · Universitas Universal, Indonesia · Beijing Language and Culture University, Beijing, China +2

LLM Evaluation LLM Alignment Moral Reasoning in Language Models Cultural Alignment Value Alignment

Jul 20, 2026cs.LG

A Geometric Perspective on Stabilizing Value Conflict Resolution

Large Language Models (LLMs) often struggle to navigate value conflicts when trained with the compressed scalar rewards of Reinforcement Learning from Human Feedback (RLHF). To address this challenge, we investigate how chain-of-thought (CoT) reasoning can help improve performance in this domain. Geometrically, we show that CoT correlates with further smoothing the model's loss landscape in its sharpest direction, helping resolve the optimization instability of traditional scalar rewards. We also demonstrate via relevant downstream benchmarks that value conflict-focused CoT may generalize to different kinds of moral reasoning, demonstrating that this CoT has the potential to be an effective mechanism for better moral reasoning. To capitalize on this potential, we create a new value conflict-focused CoT design that further smooths the sharpest direction of the loss landscape and increases moral reasoning performance. This finding shows that explicitly modifying and improving the design of reasoning dynamics offers a promising avenue for improving model performance on user requests with complex value conflicts, advancing pluralistic alignment in LLMs.

Saket Reddy, Andy Liu

Supervised Program for Alignment Research (SPAR) · University of Illinois - Urbana-Champaign · Carnegie Mellon University

LLM Alignment Moral Reasoning in Language Models CoT Reasoning

Jul 19, 2026cs.AI

Toward Anthropomorphic Dialogue: A Closed-Loop Framework for Human-Like Chat Generation, Evaluation, and Preference Alignment

Human-like private chat requires more than fluent response generation: a system must preserve persona, relationship, memory, bounded knowledge, medium-specific timing, and a coherent multi-turn arc. We present AnthroDial, a closed-loop framework that formulates anthropomorphic dialogue as a joint problem of system architecture, executable evaluation, and diagnostic alignment. It combines (1) a role-conditioned scheduled dialogue runtime with persona and scenario cards, long-term memory, virtual time, and single-draft message decisions; (2) an executable benchmark with an L0 validity gate, five per-turn dimensions, and five dialogue-level dimensions; and (3) a post-training pipeline that filters 16,436 scheduled-decision examples for SFT and applies GRPO with a cognitive-diagnostic, ZPD-aware reward. The reward maintains Kalman-filtered capability estimates for each behavioral dimension, upweights dimensions with larger capability deficits, and uses rollout scores as task-level ZPD matches to focus optimization on learnable weak skills. On a benchmark with 55 personas, 50 scenarios, 50 persona-scenario bindings, and 100 role-conditioned cases per model, we evaluate 16 systems spanning frontier baselines, open models, thinking/no-think variants, and SFT/RL ablations. The strongest non-trained baseline reaches 32.00% strict ACC, while Qwen3.6-27B-SFT+RL reaches 39.00% strict ACC and a 98.5 overall score. In the 9B no-think setting, SFT and RL improve strict ACC from 0.00% to 13.00% and 18.37%. These results show that anthropomorphic dialogue benefits when generation, evaluation, and reward shaping share the same behavioral dimensions.

Wentao Liu, Siyu Song, Xi Chen +4

1Shanghai Institute of Innovation · 2East China Normal University · University of Science and Technology of China +3

LLM Alignment Large Language Model-Based Role-Play Simulation Multi-Turn Dialogue Evaluation Conversational Agents Anthropomorphism in AI Persona-Conditioned Generation