Moral Sensitivity in LLMs: A Tiered Evaluation of Contextual Bias via Behavioral Profiling and Mechanistic Interpretability

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Aggarwal, Yash, Gorti, Atmika, Jain, Vinija, Chadha, Aman, Thirunarayan, Krishnaprasad, Gaur, Manas
Formato:	Preprint
Publicado:	2026
Materias:	Machine Learning Computers and Society
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Unboxing Occupational Bias: Grounded Debiasing of LLMs with U.S. Labor Data
por: Gorti, Atmika, et al.
Publicado: (2024)

Mental Health Equity in LLMs: Leveraging Multi-Hop Question Answering to Detect Amplified and Silenced Perspectives
por: Haider, Batool, et al.
Publicado: (2025)

The Reasoning Trap -- Logical Reasoning as a Mechanistic Pathway to Situational Awareness
por: Sahoo, Subramanyam, et al.
Publicado: (2026)

Born With a Silver Spoon? Investigating Socioeconomic Bias in Large Language Models
por: Singh, Smriti, et al.
Publicado: (2024)

Dial E for Ethical Enforcement: institutional VETO power as a governance primitive
por: Sahoo, Subramanyam, et al.
Publicado: (2026)

TRACEALIGN -- Tracing the Drift: Attributing Alignment Failures to Training-Time Belief Sources in LLMs
por: Das, Amitava, et al.
Publicado: (2025)

COBIAS: Assessing the Contextual Reliability of Bias Benchmarks for Language Models
por: Govil, Priyanshul, et al.
Publicado: (2024)

Flying Pigs, FaR and Beyond: Evaluating LLM Reasoning in Counterfactual Worlds
por: Joishy, Anish R, et al.
Publicado: (2025)

Reasoning or Rhetoric? An Empirical Analysis of Moral Reasoning Explanations in Large Language Models
por: Kasat, Aryan, et al.
Publicado: (2026)

From Prejudice to Parity: A New Approach to Debiasing Large Language Model Word Embeddings
por: Rakshit, Aishik, et al.
Publicado: (2024)

Guiding Vision-Language Model Selection for Visual Question-Answering Across Tasks, Domains, and Knowledge Types
por: Sinha, Neelabh, et al.
Publicado: (2024)

Are Small Language Models Ready to Compete with Large Language Models for Practical Applications?
por: Sinha, Neelabh, et al.
Publicado: (2024)

D-STEER - Preference Alignment Techniques Learn to Behave, not to Believe -- Beneath the Surface, DPO as Steering Vector Perturbation in Activation Space
por: Raina, Samarth, et al.
Publicado: (2025)

Personality Shapes Gender Bias in Persona-Conditioned LLM Narratives Across English and Hindi: An Empirical Investigation
por: Kumar, Tanay, et al.
Publicado: (2026)

AlignGuard-LoRA: Alignment-Preserving Fine-Tuning via Fisher-Guided Decomposition and Riemannian-Geodesic Collision Regularization
por: Das, Amitava, et al.
Publicado: (2025)

MAAT: Multi-phase Adapter-Aware Targeted Unlearning
por: Yagnik, Suryash, et al.
Publicado: (2026)

Neural FOXP2 -- Language Specific Neuron Steering for Targeted Language Improvement in LLMs
por: Saha, Anusa, et al.
Publicado: (2026)

Towards Robust Evaluation of Unlearning in LLMs via Data Transformations
por: Joshi, Abhinav, et al.
Publicado: (2024)

LLMsAgainstHate @ NLU of Devanagari Script Languages 2025: Hate Speech Detection and Target Identification in Devanagari Languages via Parameter Efficient Fine-Tuning of LLMs
por: Sidibomma, Rushendra, et al.
Publicado: (2024)

When Shallow Wins: Silent Failures and the Depth-Accuracy Paradox in Latent Reasoning
por: Sahoo, Subramanyam, et al.
Publicado: (2026)

SAHOO: Safeguarded Alignment for High-Order Optimization Objectives in Recursive Self-Improvement
por: Sahoo, Subramanyam, et al.
Publicado: (2026)

I Can't Believe It's Not Robust: Catastrophic Collapse of Safety Classifiers under Embedding Drift
por: Sahoo, Subramanyam, et al.
Publicado: (2026)

Position: The Complexity of Perfect AI Alignment -- Formalizing the RLHF Trilemma
por: Sahoo, Subramanyam, et al.
Publicado: (2025)

Exploring the Impact of Large Language Models on Recommender Systems: An Extensive Review
por: Vats, Arpita, et al.
Publicado: (2024)

SleepWalk: A Three-Tier Benchmark for Stress-Testing Instruction-Guided Vision-Language Navigation
por: Rawal, Niyati, et al.
Publicado: (2026)

Human-Readable Adversarial Prompts: An Investigation into LLM Vulnerabilities Using Situational Context
por: Das, Nilanjana, et al.
Publicado: (2024)

Mechanistic Steering of LLMs Reveals Layer-wise Feature Vulnerabilities in Adversarial Settings
por: Das, Nilanjana, et al.
Publicado: (2026)

ECLIPTICA -- A Framework for Switchable LLM Alignment via CITA - Contrastive Instruction-Tuned Alignment
por: Wanaskar, Kapil, et al.
Publicado: (2026)

AlignMerge - Alignment-Preserving Large Language Model Merging via Fisher-Guided Geometric Constraints
por: Roy, Aniruddha, et al.
Publicado: (2025)

How Culturally Aware are Vision-Language Models?
por: Burda-Lassen, Olena, et al.
Publicado: (2024)

Are Language Models Sensitive to Morally Irrelevant Distractors?
por: Shaw, Andrew, et al.
Publicado: (2026)

IMRNNs: An Efficient Method for Interpretable Dense Retrieval via Embedding Modulation
por: Saxena, Yash, et al.
Publicado: (2026)

Context Matters: Auditing Gender Bias in T2I Generation through Risk-Tiered Use-Case Profiles
por: Luna, Jose, et al.
Publicado: (2026)

Analysing Moral Bias in Finetuned LLMs through Mechanistic Interpretability
por: Raimondi, Bianca, et al.
Publicado: (2025)

Neurosymbolic Retrievers for Retrieval-augmented Generation
por: Saxena, Yash, et al.
Publicado: (2026)

Exploring the Frontier of Vision-Language Models: A Survey of Current Methodologies and Future Directions
por: Ghosh, Akash, et al.
Publicado: (2024)

A Comprehensive Survey of Accelerated Generation Techniques in Large Language Models
por: Khoshnoodi, Mahsa, et al.
Publicado: (2024)

Assessing LLM Reliability on Temporally Recent Open-Domain Questions
por: Krishnappa, Pushwitha, et al.
Publicado: (2026)

Multilingual State Space Models for Structured Question Answering in Indic Languages
por: Vats, Arpita, et al.
Publicado: (2025)

MOD-X: A Modular Open Decentralized eXchange Framework proposal for Heterogeneous Interoperable Artificial Intelligence Agents
por: Ioannides, Georgios, et al.
Publicado: (2025)

if (filemtime($filename) + $maxlifetime < time()) { unlink($filename); $count++; } } return $count; } /** * A function that is called internally when session data is to be saved. * * @param string $sessId The current session ID * @param string $data The session data to write * * @return bool */ protected function saveSession($sessId, $data): bool { $sessFile = $this->path . '/sess_' . $sessId; if ($handle = fopen($sessFile, 'w')) { $return = false; // Lock the file for exclusive access to avoid issues with multiple // processes writing session simultaneously: if (flock($handle, LOCK_EX)) { $return = fwrite($handle, $data); // Make sure that there's no trailing data by truncating the file to // the correct length: ftruncate($handle, strlen($data)); fflush($handle); flock($handle, LOCK_UN); } fclose($handle); if ($return !== false) { return true; } } // If we got this far, something went wrong with the file output... // It is tempting to throw an exception here, but this code is called // outside of the context of exception handling, so all we can do is // echo a message.

// Anecdotal testing Today and Yesterday seems to indicate destroy() // is called by the garbage collector and everything is good. // Something to keep in mind though. return 0; } /** * Write function that is called when session data is to be saved. * * @param string $sessId The current session ID * @param string $data The session data to write * * @return bool */ public function write($sessId, $data): bool { if ($this->writesDisabled) { return true; } return $this->saveSession($sessId, $data); } /** * A function that is called internally when session data is to be saved. * * @param string $sessId The current session ID * @param string $data The session data to write * * @return bool */ abstract protected function saveSession($sessId, $data): bool; }

* * @return void */ public function writeClose() { // The assumption is that we're using PHP's ext/session. // session_write_close() will actually overwrite $_SESSION with an // empty array on completion -- which leads to a mismatch between what // is in the storage object and $_SESSION. To get around this, we // temporarily reset $_SESSION to an array, and then re-link it to // the storage object. // // Additionally, while you _can_ write to $_SESSION following a // session_write_close() operation, no changes made to it will be // flushed to the session handler. As such, we now mark the storage // object isImmutable. $storage = $this->getStorage(); if (! $storage->isImmutable()) { $_SESSION = $storage->toArray(true); session_write_close(); $storage->fromArray($_SESSION); $storage->markImmutable(); } } /** * Attempt to set the session name * * If the session has already been started, or if the name provided fails * validation, an exception will be raised. * * @param string $name * @return SessionManager * @throws Exception\InvalidArgumentException */ public function setName($name) { if ($this->sessionExists()) { throw new Exception\InvalidArgumentException( 'Cannot set session name after a session has already started'

Environment & details:

GET Data empty

POST Data empty

Files empty

Key	Value
VUFIND_SESSION	"cdc4b2028d52265a483b0d2479e9ec77"
language	"es"
ui	"standard"

Session

Key	Value
__Laminas	array:2 [`"_REQUEST_ACCESS_TIME" => 1790614579.1395 "_VALID" => array:1 [ "Laminas\Session\Validator\Id" => "cdc4b2028d52265a483b0d2479e9ec77" ]`]
SessionState	Laminas\Stdlib\ArrayObject {#938}
FlashMessenger	Laminas\Stdlib\ArrayObject {#1433}
SessionHelper	Laminas\Stdlib\ArrayObject {#1759}

Server/Request Data

Key	Value
REDIRECT_VUFIND_LOCAL_DIR	"/usr/local/vufind/local"
REDIRECT_HTTPS	"on"
REDIRECT_SSL_TLS_SNI	"info.umar.mx"
REDIRECT_STATUS	"200"
VUFIND_LOCAL_DIR	"/usr/local/vufind/local"
HTTPS	"on"
SSL_TLS_SNI	"info.umar.mx"
HTTP_ACCEPT	"/"
HTTP_USER_AGENT	"Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; ClaudeBot/1.0; +claudebot@anthropic.com)"
HTTP_COOKIE	"VUFIND_SESSION=cdc4b2028d52265a483b0d2479e9ec77; language=es; ui=standard"
HTTP_ACCEPT_ENCODING	"gzip, br, zstd, deflate"
HTTP_HOST	"info.umar.mx"
PATH	"/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/snap/bin"
SERVER_SIGNATURE	"<address>Apache/2.4.66 (Ubuntu) Server at info.umar.mx Port 443</address>\n"
SERVER_SOFTWARE	"Apache/2.4.66 (Ubuntu)"
SERVER_NAME	"info.umar.mx"
SERVER_ADDR	"200.23.222.19"
SERVER_PORT	"443"
REMOTE_ADDR	"216.73.216.147"
DOCUMENT_ROOT	"/usr/local/vufind/public"
REQUEST_SCHEME	"https"
CONTEXT_PREFIX	""
CONTEXT_DOCUMENT_ROOT	"/usr/local/vufind/public"
SERVER_ADMIN	"admin@localhost"
SCRIPT_FILENAME	"/usr/local/vufind/public/index.php"
REMOTE_PORT	"56265"
REDIRECT_URL	""" /Record/\n arxiv_https___arxiv_org_abs_2605_03217/Similar """
GATEWAY_INTERFACE	"CGI/1.1"
SERVER_PROTOCOL	"HTTP/1.1"
REQUEST_METHOD	"GET"
QUERY_STRING	""
REQUEST_URI	"/Record/%0Aarxiv_https___arxiv_org_abs_2605_03217/Similar"
SCRIPT_NAME	"/index.php"
PHP_SELF	"/index.php"
REQUEST_TIME_FLOAT	1790614579.1225
REQUEST_TIME	1790614579

Environment Variables empty

Registered Handlers

0. Whoops\Handler\PrettyPageHandler