LPDS: Evaluating LLM Robustness Through Logic-Preserving Difficulty Scaling

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Mondorf, Philipp, Bell, Samuel J., Dodge, Jesse, Hupkes, Dieuwke
Formato:	Preprint
Publicado:	2026
Materias:	Machine Learning
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Compute Optimal Scaling of Skills: Knowledge vs Reasoning
por: Roberts, Nicholas, et al.
Publicado: (2025)

Quantifying Variance in Evaluation Benchmarks
por: Madaan, Lovish, et al.
Publicado: (2024)

Circuit Compositions: Exploring Modular Structures in Transformer-Based Language Models
por: Mondorf, Philipp, et al.
Publicado: (2024)

MultiLoKo: a multilingual local knowledge benchmark for LLMs spanning 31 languages
por: Hupkes, Dieuwke, et al.
Publicado: (2025)

Brittlebench: Quantifying LLM robustness via prompt sensitivity
por: Romanou, Angelika, et al.
Publicado: (2026)

ChaosBench-Logic v2: Evaluating LLM Logical Reasoning over Dynamical Systems at Scale
por: Thomas, Noel
Publicado: (2026)

From Form(s) to Meaning: Probing the Semantic Depths of Language Models Using Multisense Consistency
por: Ohmer, Xenia, et al.
Publicado: (2024)

ExPLAIND: Unifying Model, Data, and Training Attribution to Study Model Behavior
por: Eichin, Florian, et al.
Publicado: (2025)

Correlating and Predicting Human Evaluations of Language Models from Natural Language Processing Benchmarks
por: Schaeffer, Rylan, et al.
Publicado: (2025)

Reason to Rote: Rethinking Memorization in Reasoning
por: Du, Yupei, et al.
Publicado: (2025)

Holistically Evaluating the Environmental Impact of Creating Language Models
por: Morrison, Jacob, et al.
Publicado: (2025)

LLM-TabLogic: Preserving Inter-Column Logical Relationships in Synthetic Tabular Data via Prompt-Guided Latent Diffusion
por: Long, Yunbo, et al.
Publicado: (2025)

Difficulties with Evaluating a Deception Detector for AIs
por: Smith, Lewis, et al.
Publicado: (2025)

On the Difficulty of Constructing a Robust and Publicly-Detectable Watermark
por: Fairoze, Jaiden, et al.
Publicado: (2025)

Fake News Detection After LLM Laundering: Measurement and Explanation
por: Das, Rupak Kumar, et al.
Publicado: (2025)

Reasoning-as-Logic-Units: Scaling Test-Time Reasoning in Large Language Models Through Logic Unit Alignment
por: Li, Cheryl, et al.
Publicado: (2025)

Signal and Noise: A Framework for Reducing Uncertainty in Language Model Evaluation
por: Heineman, David, et al.
Publicado: (2025)

Evaluating Game Difficulty in Tetris Block Puzzle
por: Wang, Chun-Jui, et al.
Publicado: (2026)

Liar, Liar, Logical Mire: A Benchmark for Suppositional Reasoning in Large Language Models
por: Mondorf, Philipp, et al.
Publicado: (2024)

LLM Knowledge is Brittle: Truthfulness Representations Rely on Superficial Resemblance
por: Haller, Patrick, et al.
Publicado: (2025)

Lower Difficulty and Better Robustness: A Bregman Divergence Perspective for Adversarial Training
por: Wu, Zihui, et al.
Publicado: (2022)

The Pragmatic Frames of Spurious Correlations in Machine Learning: Interpreting How and Why They Matter
por: Bell, Samuel J., et al.
Publicado: (2024)

Evaluating the Robustness of Chinchilla Compute-Optimal Scaling
por: Schaeffer, Rylan, et al.
Publicado: (2025)

Data Difficulty and the Generalization--Extrapolation Tradeoff in LLM Fine-Tuning
por: Liu, Siyuan, et al.
Publicado: (2026)

Reassessing the Validity of Spurious Correlations Benchmarks
por: Bell, Samuel J., et al.
Publicado: (2024)

Limits of Difficulty Scaling: Hard Samples Yield Diminishing Returns in GRPO-Tuned SLMs
por: Yadav, Suraj, et al.
Publicado: (2026)

Compress the Easy, Explore the Hard: Difficulty-Aware Entropy Regularization for Efficient LLM Reasoning
por: Luo, Qin-Wen, et al.
Publicado: (2026)

Interpretability of Language Models via Task Spaces
por: Weber, Lucas, et al.
Publicado: (2024)

Improving Data Efficiency for LLM Reinforcement Fine-tuning Through Difficulty-targeted Online Data Selection and Rollout Replay
por: Sun, Yifan, et al.
Publicado: (2025)

Beyond Scaling Curves: Internal Dynamics of Neural Networks Through the NTK Lens
por: Nikolaou, Konstantin, et al.
Publicado: (2025)

GPAS: Accelerating Convergence of LLM Pretraining via Gradient-Preserving Activation Scaling
por: Chen, Tianhao, et al.
Publicado: (2025)

Optimizing Privacy-Preserving Primitives to Support LLM-Scale Applications
por: Jandali, Yaman, et al.
Publicado: (2025)

Circuit Transformer: A Transformer That Preserves Logical Equivalence
por: Li, Xihan, et al.
Publicado: (2024)

Scaling Up Diffusion and Flow-based XGBoost Models
por: Cresswell, Jesse C., et al.
Publicado: (2024)

Identity-Link IRT for Label-Free LLM Evaluation: Preserving Additivity in TVD-MI Scores
por: Robertson, Zachary
Publicado: (2025)

Robust Mitigation of Age-Dependent Confounding Effects via Sample-Difficulty Decorrelation
por: Kurian, Nikhil Cherian, et al.
Publicado: (2026)

D3: Diversity, Difficulty, and Dependability-Aware Data Selection for Sample-Efficient LLM Instruction Tuning
por: Zhang, Jia, et al.
Publicado: (2025)

Separable Expert Architecture: Toward Privacy-Preserving LLM Personalization via Composable Adapters and Deletable User Proxies
por: Schneider, Chris, et al.
Publicado: (2026)

Tokens-per-Parameter Coverage Is Critical for Robust LLM Scaling Law Extrapolation
por: Kricheli, Joshua Shay, et al.
Publicado: (2026)

LLM-Assisted Logic Rule Learning: Scaling Human Expertise for Time Series Anomaly Detection
por: Zhang, Haoting, et al.
Publicado: (2026)

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
language	"es"
ui	"standard"
VUFIND_SESSION	"bb900e19343dae92211ea248981b777e"

Session

Key	Value
__Laminas	array:2 [`"_REQUEST_ACCESS_TIME" => 1790870962.0142 "_VALID" => array:1 [ "Laminas\Session\Validator\Id" => "bb900e19343dae92211ea248981b777e" ]`]
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	"language=es; ui=standard; VUFIND_SESSION=bb900e19343dae92211ea248981b777e"
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	"14386"
REDIRECT_URL	""" /Record/\n arxiv_https___arxiv_org_abs_2605_15393/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_15393/Similar"
SCRIPT_NAME	"/index.php"
PHP_SELF	"/index.php"
REQUEST_TIME_FLOAT	1790870962.0072
REQUEST_TIME	1790870962

Environment Variables empty

Registered Handlers

0. Whoops\Handler\PrettyPageHandler