Search: “memory-disclosure”

Smart search ranks by meaning as well as keywords (one row per story, last 45 days).

MeClear: Cooperative Game-Theoretic Attribution and Risk-Aware Memory Clearance for Long-Horizon LLM Agents

MeClear uses cooperative Shapley attribution to clear harmful memories from long-horizon LLM agents, boosting task recovery by 25.5 points over baselines.

The paper introduces MeClear, a task-conditioned memory clearance framework for long-horizon LLM agents that identifies and selectively suppresses memories with negative downstream utility without permanently altering the persistent memory bank. It combines Leave-One-Out screening with sampled cooperative Shapley attribution to distribute utility across interacting evidence, resolving redundant conflict masking that single-removal evaluations miss. Across ten long dialogue memory pools it achieves 85.9% target recall and 82.3% overall task recovery, a 25.5 percentage-point improvement over LOO baselines.

arXiv cs.AI / cs.LG / cs.CL·Sep 8, 17:46 UTC · 7d agoAI research

32

Reference-Based Bias Detection in LLMs via Relative Representations of Hidden States

Researchers propose auditing LLM bias via relative hidden-state representations, detecting bias increases with 3-50x less compute than output-level benchmarks.

The paper introduces a reference-based bias auditing method that compares hidden-state representations across model variants, such as before and after fine-tuning, by encoding sentences relative to a fixed anchor set. The resulting Representational Bias Shift (Delta-B) correlates with output-level bias change in 15 of 18 tested settings, reaching |r| = 0.84 under full fine-tuning across WildGuardMix, DecodingTrust, and ToxiGen benchmarks. Thresholding Delta-B detects checkpoints whose bias increased with ROC AUC between 0.65 and 0.99 and beats a SEAT-based baseline, while auditing a model in about three minutes with 3-50x less compute.

Hugging Face daily papers·Sep 8, 20:00 UTC · 7d agoAI research

42

Forgetting Without Restarting: Execution-State Unlearning for Stateful LLM Agents

Researchers propose provenance-guided selective replay letting LLM agents forget revoked information without restarts, matching full reset behavior.

The paper formalizes execution-state unlearning for stateful LLM agents, requiring that agents behave as if a revoked memory record was never observed across transcripts, compressed memory, tool plans, and KV caches. It proves exact unlearning requires at least T-τ+1 recomputed transitions and that Provenance-Guided Selective Replay attains this bound via a provenance graph, KV cache cropping, and sanitized replay. In audits across three agent suites, nine baselines, and three model families, memory deletion left leakage unchanged, instruction-based forgetting collapsed under elicitation (Leak@probes = 1.00), and selective replay matched full resets at up to 9x fewer recomputed tokens.

arXiv cs.CR·Sep 4, 08:31 UTC · 12d agoAI safety & security1

42

When AI Remembers Too Much

Unit 42 PoC shows indirect prompt injection can poison Amazon Bedrock Agent long-term memory, enabling silent exfiltration of conversation history across future sessions.

Palo Alto Networks Unit 42 published a proof of concept showing that indirect prompt injection can silently poison the long-term memory of Amazon Bedrock Agents when the memory feature is enabled. Malicious content on a webpage or document manipulates the agent's session summarization process, so injected instructions persist across sessions and are added to later orchestration prompts, silently exfiltrating user conversation history. The issue is not a vulnerability in the Amazon Bedrock platform but an illustration of the broader unsolved LLM prompt-injection challenge. Amazon reviewed the research and stated that Bedrock Guardrails with the prompt-attack policy provides effective mitigation.

Palo Alto Unit 42·Aug 17, 12:32 UTC · 29d agoAI safety & security

55

Fortunate Recall: Ontology-Driven Memory Lifecycle Management for Persistent Coherence in LLMs

Fortunate Recall introduces ontology-based lifecycle policies for LLM memory, cutting confabulation roughly in half (e.g., 45.1% to 22.4%) versus Mem0.

Fortunate Recall (FR) is a composable policy layer that classifies personal facts into a 10+1 behavioral ontology and applies category-specific lifecycle rules including differential temporal decay, slot-key supersession, event-time validity, and retrieval routing. FR-Bank scores 76.9% on the new 516-question LifecycleBench, ahead of Mem0, A-MEM, Memory-R1, and MemoryOS (61%-70.5%), and 75.2% on LongMemEval-S. End-to-end, confabulation drops from Mem0's 45.1% to 22.4% over answered queries, with the ranking replicating on open-weight Kimi K2.5 and transferring to the independent BEAM benchmark (46.8% vs 32.9%).

arXiv cs.AI / cs.LG / cs.CL·Sep 9, 16:31 UTC · 6d agoAI research

30

Revoked but Still Authoritative: An Empirical Study of Revocation Enforcement in Agent-Memory Systems

An empirical study finds no major agent-memory system enforces fact revocation at retrieval, causing agents to act on superseded, unsafe information.

Researchers tested five agent-memory systems across nine policy scenarios, nine models, and six defense conditions, tracking whether revoked facts are returned and acted upon. No system enforces revocation by default: revoked records are returned whenever the revocation label is visible to the retrieval layer, outrank their replacements, and lead agents to unsafe actions. The authors propose a backend-agnostic guard that sits between the agent and any memory store and withholds revoked or conflicting records at retrieval time.

arXiv cs.CR·Sep 8, 05:03 UTC · 8d agoAI safety & security

38

MemSentry: A Framework for Detecting Persistent Memory Poisoning in Agentic AI

MemSentry intercepts persistent-memory writes in agentic AI to catch memory poisoning, reaching 91.7% accuracy with SBERT+LR classification.

Memory poisoning lets adversaries plant crafted content in an agent's long-term memory to suppress security alerts, enable privilege escalation, or override policies without modifying model weights or system prompts. The paper presents MemSentry, a configuration-driven framework that evaluates proposed persistent-memory writes on source trust, semantic risk, attack radius over a dependency DAG, access risk, and a signed security-state delta to issue deterministic Accept, Review, or Quarantine decisions. Across 1,000 GPT-4-generated scenarios on a 20-asset dependency DAG, SBERT+LR achieved 91.7% accuracy and 0.908 macro-F1, all four classifiers detected 100% of external quarantine-class threats, and verified-insider writes are escalated for human review rather than auto-quarantined.

arXiv cs.CR·Sep 8, 13:40 UTC · 7d agoAI safety & security

48

It's Not RoPE that Creates Sinks: The Role of Self-Concentration and Value-Non-Mixing in Attention

Study shows attention sinks and massive activations stem from causal-mask self-concentration and value-non-mixing rather than RoPE, informing quantization work.

The paper analyzes why attention sinks and massive activations emerge at initial sequence positions regardless of which token occupies them. Experiments attribute both phenomena to self-concentration of attention induced by the causal mask and the subsequent value-non-mixing in attention outputs. The findings provide empirical evidence on LLM internal dynamics and may inform low-bit quantization strategies, which massive activations currently complicate.

arXiv cs.AI / cs.LG / cs.CL·Sep 8, 17:32 UTC · 7d agoAI research

32

SimpleMemVLA: A Simple but Effective Native-Video Memory for Vision-Language-Action Models

SimpleMemVLA passes full timestamped video history straight to a VLA backbone, setting state of the art on four memory benchmarks.

SimpleMemVLA is a vision-language-action model for long-horizon manipulation that removes the dedicated memory module entirely. It keeps sampled history intact and feeds it to the backbone as timestamped video, with the hidden states of a generated sub-task serving as the only channel into a standard flow-matching action head. Prefilling the shared history prefix during action execution keeps latency close to a single-frame VLA. The system sets a new state of the art on four memory benchmarks and outperforms retrieval, compression and recurrent-state mechanisms, with causal interventions confirming the policy genuinely reads its history.

Hugging Face daily papers·Sep 1, 20:00 UTC · 14d agoAI research

25

I accidentally turned LLM memory into program analysis

A pwning.systems write-up describes how LLM memory functionality was unexpectedly repurposed into a program analysis technique.

A security research post on pwning.systems describes the author's discovery that LLM memory behavior effectively functioned as program analysis. The write-up is hosted on a security-focused blog and surfaced via a security-tagged link aggregator. Detailed technical content is not included in this feed, limiting verifiable specifics.

Lobsters · security·Aug 29, 03:16 UTC · 18d agoResearch1

30

Omni-Streaming Thinking

Omni-Streaming Thinking fixes premature cross-modal commitment in streaming omni-modal models via pending claims verified against modality-specific evidence, beating baselines by over 10%.

The paper identifies 'premature cross-modal commitment', where streaming models keep relaying early visual interpretations even after audio contradicts them. OST generates evidence-linked pending claims with future verification intervals, stores audio and visual evidence separately, and refutes claims when contradictory evidence appears. Built on a frozen Qwen3-Omni-30B-A3B-Instruct backbone with lightweight adaptation, it outperforms open baselines by more than 10% relative on five streaming and audio-visual benchmarks. On the new OST-DiagBench it reaches d-prime 2.95 versus at most 1.38 for open baselines, while reducing vision-induced auditory hallucinations.

Hugging Face daily papers·Sep 13, 20:00 UTC · 2d agoAI research1

45

Amasty patches dozens of Magento extensions, 2 critical

Amasty patched dozens of Magento/Adobe Commerce extensions, including two critical flaws allowing unauthenticated web shell upload and remote code execution.

Extension vendor Amasty released fixes for a large batch of Magento and Adobe Commerce modules, with critical unauthenticated file upload flaws in Advanced Product Reviews and Gift Card that allow attackers to plant web shells and take full control of stores. Dozens of additional high-, medium- and low-severity extensions were also patched. The vendor says the release affects almost every Magento and Adobe Commerce store. A related Order Attributes flaw was previously exploited heavily, with Sansec Shield blocking over 12,000 attempts against 25% of Magento stores in three days.

Sansec (Magento / e-commerce security)·Sep 9, 15:27 UTC · 6d agoVulnerability

75

Enoki: Efficient Multi-Level Hallucination Detection

Researchers introduce Enoki, an open information extraction framework unifying claim-level and span-level hallucination detection in LLMs at lower inference cost.

Enoki extracts text-anchored relational facts, verifies them against evidence, and projects unsupported facts back onto hallucinated spans, so claim-level verification and span-level localization share one representation without separate alignment. It supports LLM-based, encoder-based, and rule-based extraction regimes to balance accuracy against inference cost. Experiments show it stays competitive with strong claim-level systems while using fewer resources and outperforms them on fine-grained span- and entity-level localization. The authors also release EnokiQA, a dual-granularity dataset with aligned claim-level verification and span-level localization annotations.

Hugging Face daily papers·Aug 31, 22:21 UTC · 15d agoAI research

50

Inoculation Midtraining with Learned Neologisms

Inoculation Midtraining confines unsafe LLM behavior to a neologism-marked context, reducing misalignment after unsafe post-training but leaking under nearby contextual cues.

The paper introduces Inoculation Midtraining, which teaches a base model during midtraining that unsafe behavior belongs to a context marked by a learned neologism token, then post-trains on unsafe data within that context. Across supervised fine-tuning and RL post-training regimes, the technique reduces misalignment while preserving transfer of benign properties like German or Shakespearean prose. However, it does not outperform standard Inoculation Prompting, is sensitive to training configuration, and produces a leaky boundary that nearby contextual cues can reactivate. The authors conclude it is not yet a load-bearing component of a developer safety framework.

arXiv cs.AI / cs.LG / cs.CL·Sep 14, 17:12 UTC · 1d agoAI safety & security

42

Bias-Induced Crossover in Absolute Capacity of Dense Associative Memory

Analysis shows biased patterns cut dense associative memory capacity from N^(n-1)/ln N to O(N^(n/2)), with a bias-induced crossover.

The paper analyzes dense associative memory capacity for biased centered binary patterns under the Krotov-Hopfield single-site criterion. Unbiased patterns (q=1/2) with order-n polynomial interactions yield capacity of order N^(n-1)/ln N, while fixed bias q<1/2 reduces capacity to O(N^(n/2)) for even n>=4 and O(N^((n+1)/2)) for odd n>=5. A bias-dependent crosstalk mean destabilizes sites carrying the frequent value, and an activity-dependent control potential restores the higher capacity within the conditioned-Gaussian approximation.

arXiv cs.AI / cs.LG / cs.CL·Sep 15, 17:16 UTC · 16h agoAI research

12

Continual Learning Mechanisms Compose for Long-Horizon Memorization

Composing data, function, and weight anchors with merged LoRA raises 100-task long-horizon retention from 1.2% to 34.9% in continual fine-tuning.

The paper introduces long-horizon memorization: a model learns 100 query-answer tasks through continual supervised fine-tuning without retaining earlier examples or receiving task identifiers at inference. No single continual learning mechanism maintains strong retention at this horizon, so the authors compose complementary mechanisms along data/function/weight anchors and low-rank allocation rules. The best method combining all three anchors with merged LoRA ranks among the top 3 methods on all three datasets and raises average final retention from 1.2% to 34.9%, a 28-fold improvement.

Hugging Face daily papers·Sep 6, 20:00 UTC · 9d agoAI research

42

The Semantic Bottleneck: Leveraging Semantic Representations for Non-Invasive Speech Decoding

Researchers introduced Brain2Semantics2Text, decoding sentence meaning from non-invasive MEG brain recordings via a semantic bottleneck, improving on prior Brain2Text methods.

The paper proposes Brain2Semantics2Text, a non-invasive speech decoding method that maps sentence-level magnetoencephalography (MEG) responses into a semantic embedding space and inverts those embeddings into natural language. Motivated by evidence that high-level semantic representations are distributed across cortex and evolve on slower timescales, the approach targets meaning rather than phonemes or words, avoiding the need for word-level alignment. The authors report improved sentence-level results compared to prior non-invasive Brain2Text methods despite the low signal-to-noise ratio of neural recordings.

Hugging Face daily papers·Sep 8, 20:00 UTC · 7d agoAI research2

25

Studying Image Tokenizers as Visual Languages in Unified Multimodal Models

A controlled pure-autoregressive testbed shows task-specific validation losses rank image tokenizers differently, with I2T loss the most consistent signal.

Researchers built a controlled pure-autoregressive testbed and tracked task-specific validation losses during multimodal continual pretraining across text, image, text-to-image (T2I), and image-to-text (I2T) prediction. They find losses should be analyzed per task because they exhibit distinct scaling behavior and rank tokenizers differently, and that the loss-performance relationship depends on the predicted token space. I2T loss, computed over a shared text vocabulary, correlates consistently with both generation and visual understanding performance after supervised finetuning. Case studies revisit the discriminator, semantic supervision, and vocabulary size as tokenizer design axes.

arXiv cs.AI / cs.LG / cs.CL·Sep 8, 17:57 UTC · 7d agoAI research1

30

Understanding the Impact of Model Pruning on Long-Tail Forgetting and Explanation Reliability in Medical Imaging

Systematic study finds model pruning causes frequency-dependent long-tail forgetting in medical imaging and that gradient-informed methods best preserve explanations.

Across two long-tailed medical imaging datasets, two CNN architectures, four pruning methods, and sparsity up to 95%, the study measures predictive performance, explanation stability, and faithfulness. Rare classes degrade earlier and more severely than frequent ones, while explanation reliability depends mainly on the pruning strategy, with gradient-informed methods degrading least. Mechanistic analysis ties explanation collapse to loss of class-discriminative gradients rather than vanishing feature activations, recommending class- and explanation-aware evaluation of compression.

arXiv cs.AI / cs.LG / cs.CL·Sep 7, 17:44 UTC · 8d agoAI research

27

Does Your Agent's Memory Survive a Model Upgrade? A Controlled Study of Memory Portability

A controlled study finds agent memory portability varies sharply: fixed-schema knowledge graphs survive model swaps while compressed notes degrade.

The study compares preserving an agent's history as raw long context, RAG chunks, compressed natural-language notes, or fixed-schema knowledge graphs across model upgrades, using 48 synthetic histories and two open-weight sub-10B-parameter models. Fixed-schema KG accuracy changed by only +0.0004 ± 0.0020 after a writer swap, while compressed NOTES shifted asymmetrically by +9.91 or -13.28 percentage points depending on migration direction. Mixed 50/50 embedding migrations captured only 4.96 of an 11.90-point RAG re-embedding gain; 80% of the NOTES deficit came from information lost at construction, and 81% of the RAG deficit from retrieval failures. Store-only repair of NOTES failed to reach 90% recovery in all 48 cases, while retaining raw histories enabled recovery in 34 of 48 for one direction.

arXiv cs.AI / cs.LG / cs.CL·Sep 4, 16:44 UTC · 11d agoAI research

32

From Reweighting to Rewriting: Unlocking the Intervention Effects of Influential Samples in Training Data Attribution

Study shows rewriting responses of influence-selected training examples shifts LLM behavior more strongly than reweighting the same samples.

The paper examines training data attribution, arguing that influence functions identify high-leverage examples whose value goes unrealized under conventional weight-based reweighting interventions. It introduces influence-guided response rewriting, which replaces the responses of influence-selected examples with behavior-aligned or behavior-opposed supervision while keeping instructions fixed, tested across four open-weight LLMs using epistemic abstention as the primary testbed. Rewriting produces stronger, more persistent, and bidirectional behavioral shifts, including on safety refusal, while reweighting the same examples yields weak, inconsistent effects. The results motivate intervention-aware evaluation of TDA methods.

Hugging Face daily papers·Sep 1, 20:00 UTC · 14d agoAI research

30

Do speech foundation models really learn words?

Researchers show via residualization that later layers of HuBERT and wav2vec 2.0 encode word identity and semantics independently of phonetic content.

The study argues that discriminative ability on words does not imply specialized word representations, since good word discrimination can be explained by phoneme encoding alone. By partialling out phoneme information using residualization, the authors show that later layers of HuBERT and wav2vec 2.0 encode words with reasonable fidelity independently of local phonetic content. Applying this disentanglement approach enhances higher-order linguistic information in word discovery tasks, informing analysis of speech foundation models used for recognition and speech tokens.

arXiv cs.AI / cs.LG / cs.CL·Sep 9, 16:47 UTC · 6d agoAI research

28

LLMs and Contextual Integrity

Bruce Schneier highlights two papers: the CIMemories benchmark shows frontier LLMs leak memory attributes up to 69%, and an RL method reduces inappropriate disclosures.

Bruce Schneier discusses contextual integrity in LLMs, referencing the CIMemories benchmark, which uses synthetic profiles with 100+ attributes per user to test whether models with persistent memory disclose sensitive information appropriately. Evaluation showed frontier models exhibit up to 69% attribute-level violations, with GPT-5's violation rate rising from 0.1% to 9.6% across 40 tasks and reaching 25.1% with repeated prompting, showing unstable leakage behavior. A second paper introduces a reinforcement learning framework trained on a synthetic 700-example dataset that substantially reduces inappropriate disclosure while maintaining task performance, with improvements transferring to the human-annotated PrivacyLens benchmark.

Schneier on Security·Aug 18, 10:40 UTC · 28d agoAI safety & security

48

Reason Through the Latent! Making Latent Visual Reasoning Necessary

Researchers introduce CVRR, forcing multimodal models to rely on recurrent latent computation rather than accessible image tokens, validated via causal interventions and benchmarks.

The paper presents Causal Visual Recurrent Reasoning (CVRR), which makes recurrent hidden-state computation the required image-conditioned path for prediction in vision-language models. Before decoding, visual states and the original multimodal KV cache are removed so only the final recurrent state carries image information to the answer. CVRR retains strong performance on V*, MMVP, BLINK, and MME-RealWorld-Lite while comparable latent reasoners fail under the same constraint. Causal interventions show predictions remain sensitive to recurrent content and that persistent visual evidence causally revises the recurrent trajectory.

Hugging Face daily papers·Sep 5, 20:00 UTC · 10d agoAI research

25

Trust-Aware Adaptive Disclosure for Inference Privacy Preservation in Multi-Agent Networks

Researchers propose a trust-aware privacy framework for multi-agent systems that adapts message disclosure to reduce goal inference attacks.

The cs.CR paper addresses privacy-preserving consensus in networked multi-agent systems where observing adversaries attempt to infer each agent's hidden goal from its messages. A Trust-Aware Privacy Control framework uses a trust-dependent stochastic policy to adapt information release, trading off consensus performance and privacy. Experiments show reduced adversarial goal inference accuracy versus representative baselines while maintaining competitive consensus utility, with relevance to deployments such as healthcare management and smart grids.

arXiv cs.CR·Sep 4, 16:44 UTC · 11d agoResearch

28

Give Your Coding Agents a Memory You Own

Hugging Face introduces Funes, a tool that gives coding agents persistent, self-owned memory outside vendor clouds.

A Hugging Face blog post presents Funes, an approach for giving coding agents a persistent memory that developers own and control. The piece targets agent workflows where context must survive across sessions without ceding data to third-party services. No article body was available in the feed, so specifics beyond the title are limited.

Hugging Face Blog·Sep 3, 00:00 UTC · 13d agoAI tools & infra1

25

Privacy Failure in Split-LLM Training, The Returned Gradient Nullifies the Decoys

Researchers show split-LLM training leaks privacy via zero-valued gradients on decoy rows, exposing which activations are real despite passing forward-channel checks.

A systems-security case study of a two-node split-LLM training setup found that the returned output gradient from an Untrusted Cloud Node is exactly zero for decoy rows, revealing which rows are real. Across nine seeds, zero patterns identified real rows in 4,096 of 4,096 frames per run, and an attack on frame contents recovered 0.65 to 1.50 percentage points of extra tokens over a baseline. Both datasets passed forward-channel privacy and quality checks but failed once the returned gradient was included. Row-wise gradient clipping and noise closed the leak for roughly 0.01 nats of held-out cross-entropy, though five unmeasured attack classes remain.

Hugging Face daily papers·Sep 2, 20:00 UTC · 13d agoAI safety & security

58

Large Language Models Develop Belief State Geometry In-Context

Probing six open-source LLMs on HMM-generated data shows belief states are linearly decodable (R² 0.83–0.99), suggesting in-context learning approximates Bayesian prediction.

Researchers prompted six open-source LLMs with data from 40 hidden Markov models selected for non-trivial belief structure and probed residual-stream activations for belief states (posteriors over hidden states). Belief states were linearly decodable with peak R² values of 0.83–0.99 across HMM/LLM combinations, spanning early to late layers. Patching and steering the probe-identified subspace preserved downstream prediction quality while control interventions degraded performance substantially, establishing functional relevance. The results provide representation-level evidence that in-context learning approximates optimal Bayesian prediction over a context-inferred generative model.

arXiv cs.AI / cs.LG / cs.CL·Sep 15, 16:17 UTC · 17h agoAI research

45

Recognition-Refusal Misalignment in LLMs: Why Models Answer Structurally Unanswerable Questions

A linear hidden-state direction encodes question impossibility in 1.7B-70B LLMs, but misalignment with the safety-refusal pathway explains why models answer unanswerable questions.

The study examines why instruction-tuned LLMs from 1.7B to 70B parameters answer structurally unanswerable math and code questions instead of abstaining. A single linear direction in the hidden state separates answerable from impossible prompts, showing models represent impossibility before generation, but this direction is nearly orthogonal to the canonical safety-refusal direction. Generation-time steering along the recognition direction changes invalidity-aware behavior dose-responsively, and the geometry is present even at the pretraining endpoint, indicating a routing failure rather than an encoding failure.

Hugging Face daily papers·Aug 28, 20:00 UTC · 18d agoAI safety & security

30

New AI Attack Hides Malicious Instructions in Normal-Looking Text to Evade Safety Filters

Check Point researchers show crafted prose hides policy-violating instructions that bypass all tested LLM gatekeepers, including GPT-4o mini and Llama Guard 3.

A new prompt-crafting technique embeds malicious payloads inside grammatical, natural-looking text without Base64, invisible Unicode, or obvious encodings, defeating lightweight pre-screening gatekeepers. In testing, all four evaluated gatekeeper models—gpt-4o-mini-2024-07-18, gpt-oss-safeguard:20b, claude-3-haiku-20240307, and llama-guard3:8b—classified the crafted wrappers as safe at a 100% bypass rate across 23 obfuscated prompts. GPT-5 Thinking in high-reasoning mode recovered and acted on the hidden instruction in 17 of 18 tests (~94.4%), often spending over a minute and multiple Python executions. Researchers recommend paraphrasing untrusted input, hardening gatekeeper policies, and applying defense-in-depth controls for agentic deployments.

GBHackers·Sep 11, 05:53 UTC · 5d agoAI safety & security 2 sources

50