Search: “Arup”

Expert-Space Exploration in MoE Reinforcement Learning

ESRL explores MoE expert-routing space during RL, lifting Qwen3-30B-A3B Pass@1 by 3.2 points over GRPO at no extra cost.

ESRL (Expert-Space Exploration Reinforcement Learning) is an architecture-aware framework that treats expert routing in Mixture-of-Experts models as an additional source of rollout diversity. It preserves high-confidence experts as anchors, restricts stochastic routing to a plausible candidate pool, adapts perturbation strength to router entropy, and replays recorded expert paths during policy optimization. ESRL achieves the best performance across top-K, top-1, and shared-expert routing backbones on mathematics, science, and code tasks; on Qwen3-30B-A3B it improves average Pass@1 and Pass@8 over GRPO by 3.2 and 4.5 percentage points respectively, without additional sampling or compute.

arXiv cs.AI / cs.LG / cs.CL·Sep 11, 16:58 UTC · 5d agoAI research 2 sources

Phishing 3.0: The Fight Moves to Agent Versus Agent

Agentic AI transforms phishing economics, enabling personalized multi-channel attacks with deepfakes like the $25M Arup deepfake heist.

The article argues phishing has evolved through three stages: from malicious content, to intent-based BEC, to AI-powered multi-channel campaigns where attacker agents autonomously conduct reconnaissance and generate tailored lures. The widely reported Arup case saw a deepfake video call impersonating colleagues convince an employee to approve transfers worth roughly $25 million. An Osterman Research study of 128 security leaders found 88% experienced trust-undermining incidents, while Microsoft 365 EOP and Google Workspace were measured missing hundreds of phishing messages per 100 mailboxes monthly. The author argues defenders must adopt their own agents to match attacker speed.

The Hacker News·Aug 19, 11:30 UTC · 28d agoPhishing & fraud2

PEEP: A Browser RAT Posing as a Chrome Extension

SOCRadar's STRU analyzed PEEP, an emerging Chromium-based RAT disguised as a Chrome extension enabling post-exploitation control of browsers.

SOCRadar's Threat Research Unit (STRU) identified and analyzed PEEP, a Chromium-based post-exploitation toolkit distributed as a browser extension. PEEP provides RAT-style capabilities while posing as a legitimate Chrome extension. It is assessed as an emerging threat worth monitoring by defenders.

SOCRadar·Sep 4, 09:50 UTC · 12d agoMalware

FulcrumSec Claims Responsibility for Manchester Airport Group Breach

FulcrumSec leaked ~549GB of Manchester Airport Group data, claiming 8.7M customer profiles exposed via exposed Iterable admin keys.

FulcrumSec posted around 549GB of uncompressed stolen Manchester Airport Group (MAG) data on its leak site, claiming nearly 8.7 million customer profiles with email, name, phone, home town, postcode and residential IP. The group said initial access came from Iterable platform admin keys exposed in the root-domain JavaScript of the Manchester, Stansted and East Midlands airport websites. Allegedly stolen data also includes ~1.2 billion marketing events, 2.5 million bookings, 461,000 SMS records, 108,000 vehicle plates and ~191,000 future bookings. MAG has provided no update since August 27 and the claims remain unverified.

Infosecurity Magazine·Sep 2, 08:15 UTC · 14d agoData breach

PACE: Perceived-Latency-Aware Cascading Service Routing and Filler Control for QoE-Efficient Retrieval-Augmented Dialogue Serving

PACE cuts perceived latency in retrieval-augmented dialogue serving via cascading routing and filler control, reaching 0.41s P95 under load.

PACE is a serving framework for retrieval-augmented dialogue that optimizes Perceived Time-to-First-Response (PTFR) under quality and cost constraints. It combines a load-adaptive cascading router, a joint path-filler controller, and volatility-aware cache admission, deployed on a humanoid-robot sales service. On 75k CarQA requests, the cascade halved pure-LLM P95 PTFR (0.29s vs 0.53s) and the adaptive controller reached 0.41s P95, 2.4x better than RAG at high load; filler calls dropped 94% and stale answers fell from 86% to 0%.

arXiv cs.AI / cs.LG / cs.CL·Sep 9, 16:02 UTC · 7d agoAI tools & infra

PIDS-Bench: Evaluating Prompt-Injection Detectors Under Over-Defense, Obfuscation, and Distribution Shift

PIDS-Bench shows prompt-injection detectors scoring F1 above 0.98 still misclassify about one-third of external benign security-adjacent prompts, revealing provenance-sensitive over-defense.

PIDS-Bench is a frozen multi-axis benchmark that jointly evaluates prompt-injection detectors on attack detection and benign false-positive behavior at fixed thresholds, spanning in-distribution inputs, hard-benign prompts, obfuscated attacks, and domain/structural distribution shifts. It evaluates seven detectors plus a rule-based lower-bound reference. A detector exceeding F1 = 0.98 on held-out data still misclassifies roughly one-third of an externally-sourced benign security-adjacent subset, and no internal detector reaches F1 >= 0.95 with hard-benign FPR <= 0.10 on the stress distribution. Hard-negative augmentation nearly eliminates over-defense on curated stress inputs but leaves it intact on externally-sourced prompts, a pattern termed provenance-sensitive over-defense.

arXiv cs.CR·Sep 14, 04:28 UTC · 2d agoAI safety & security

Why Is SHAP Not a Reliable Standalone Explanation Framework for Malware Detection?

arXiv paper shows SHAP gives unreliable standalone explanations for malware detection, with attribution dilution and sign reversal in dependent PE feature spaces.

The paper argues SHAP's formal guarantees are insufficient for reliable malware interpretation because the explained feature-coalition game is fixed only by analyst choices, not by malware behavior in the data. In static Portable Executable feature spaces, dependent feature groups cause conditional SHAP to dilute credit by a factor of 1/m across redundant features, attribute importance to features the model never uses, and even reverse attribution signs; interventional SHAP queries off-manifold coalitions no real executable exhibits. Experiments on EMBER-2018, EMBER-2024, and BODMAS with fixed LightGBM and XGBoost detectors confirm these effects. The authors position SHAP as a limited diagnostic requiring explicit data-distribution statements and domain validation, not a standalone explanation framework.

arXiv cs.CR·Sep 4, 02:00 UTC · 12d agoResearch1

PEEP Turns Chrome and Edge Into Post-Compromise Backdoors for Host Command Execution

SOCRadar disclosed PEEP, a post-exploitation Chromium extension toolkit that turns Chrome and Edge into host-level backdoors via native messaging and C2 polling.

SOCRadar researchers detailed PEEP, a Chromium-based post-exploitation toolkit masquerading as a 'Smart Bookmarks' extension injected directly into Chrome and Edge profiles by forging Secure Preferences integrity values, bypassing Web Store checks. A native-messaging host binary (nm_host.exe) extends the browser agent to host-level command execution, file management, credential theft, and session hijacking, while the extension polls C2 endpoints at 206.237.30.232 or xfjcc.fun every 30 seconds over plaintext HTTP and exfiltrates browsing history, cookies, and tab metadata. PEEP is derived from the open-source RedExt red teaming framework, adds PowerShell persistence scripts, and includes a Linux-targeting Python script, suggesting cross-platform intent. The activity is unattributed but contains Chinese-language artifacts; a C2 /health endpoint showed 34 agent entries and 10 active sessions.

The Hacker News·Sep 7, 18:12 UTC · 9d agoMalware in the wild1

CiteShade: Citation Laundering in Multi-Source Retrieval-Augmented Generation and Its Counterfactual Defense

CiteShade attack makes RAG models cite trusted sources for attacker-chosen wrong answers, raising wrong-answer rate from 0.01 to 0.68.

CiteShade is presented as the first citation laundering attack against multi-source retrieval-augmented generation: an attacker controlling a single source induces a wrong answer falsely attributed to a trusted source, even while correct evidence remains in context. The attack is formalized via three necessary conditions (retrieval, generation, citation) constructible without any instructions, raising wrong-answer rate from 0.01 to 0.68 on multi-hop QA, with source deletion confirming the malicious source as causal driver. Vulnerability tracks a model's citation propensity rather than scale, reaching CLR 0.84 with explicit instruction and 0.64 without on the most citation-prone model. Perplexity filtering and citation-support checking prove insufficient; the authors propose a counterfactual defense verifying which source actually drove the answer.

arXiv cs.CR·Sep 14, 14:41 UTC · 2d agoAI safety & security1

From Protocols to Evidence: Bounded Claims for AI in Service of the Common Good

Paper proposes a rupture test and RISE AI architecture for evidence-bounded responsible-AI claims, framed via EU AI Act and NIST AI RMF.

The paper argues AI deployment intervenes in pre-existing institutional failures of responsiveness, belonging, care, and accountability, and must therefore evaluate both the system and the institutional rupture it enters. It reviews how the EU AI Act, NIST AI RMF, and ISO/IEC 42001 translate principles into protocols, and draws on Pope Leo XIV's Magnifica Humanitas to develop a rupture test linking institutional baselines to system evaluation. It distinguishes evidence-bounded deployment from measurement-bounded governance and introduces RISE AI, an architecture for bounded claims about Responsibility, Inclusivity, Safety, and Empowerment.

arXiv cs.AI / cs.LG / cs.CL·Sep 10, 17:56 UTC · 6d agoAI policy

ActReview: Rebuttal-Guided Training Data and Rubric Rewards for Actionable Peer Review Generation

ActReview post-trains Qwen3-8B-Base on OpenReview rebuttals to generate actionable peer-review feedback with grounded revision suggestions, benchmarked on 1,000 curated instances.

The paper defines Actionable Peer-review Generation as diagnostic claim generation plus revision suggestion generation and introduces ActReview, a rebuttal-guided post-training framework. From OpenReview review-rebuttal threads the authors build ActReview-40K, aligning reviewer weaknesses with author responses grounded in localized paper evidence, and post-train Qwen3-8B-Base with multi-task SFT followed by GRPO using weakness-specific rubric rewards. They also release ActReview-Bench, a human-curated 1,000-instance benchmark, on which ActReview outperforms prior specialized review-generation models on actionability and grounding while remaining competitive with strong prompt-based LLMs. Human evaluation confirms improved revision usefulness but identifies a remaining gap in technical accuracy.

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

SkillSpector: NVIDIA's open-source security scanner for AI agent skills

Help Net Security·Aug 3, 00:00 UTC · Aug 3, 2026Malware

Distill Globally, Adapt Locally: Reasoning Distillation and Product-Type Test-Time Training for Scalable Trade-Up Recommendation

A distillation framework compresses LLM reasoning into a 15.5M-parameter trade-up recommendation model reaching AUC 0.941 with product-type test-time training.

The paper targets trade-up recommendation, which identifies higher-quality alternatives that preserve customer purchase intent. A retrieval-augmented few-shot LLM teacher generates labels and rationales that supervise a compact embedding-pair classifier; at inference the 15.5M-parameter student uses only two precomputed 768-dimensional embeddings with no LLM calls. On 8,352 annotated pairs, label-only training scored AUC 0.912, reasoning distillation reached 0.924, and product-type test-time training lifted it to 0.941 with average precision 0.940. The distilled student is roughly 5,000x faster and 10,000x cheaper than direct LLM inference on a 100K-pair proxy catalog.

arXiv cs.AI / cs.LG / cs.CL·Sep 4, 17:08 UTC · 12d agoAI research

Import AI 470: No rights for machines; automating environment generation with SPADE; and building better GPU kernels with Hawkeye

METR analysis finds AI accelerating cyber vulnerability discovery, while SPADE self-play environment generation improves Qwen3 reasoning benchmark scores at 30B scale.

Import AI 470 discusses a METR research note reporting differential acceleration from AI: major acceleration in reported cyber vulnerabilities (cURL, OpenSSL, Firefox, Microsoft, NVD, OSV), minor acceleration in mathematics, and no measurable acceleration in AI-research optimization benchmarks. It also covers SPADE, a self-play framework from a multi-university team (University of Washington, Stanford, MIT, CMU, and others) that co-evolves executable training environments and agent capability using Environment Designer and Reasoning Agent roles with hint-based regret rewards. Trained on Qwen3-4B-Instruct-2507, Qwen3-8B, and Qwen3-30B-A3B-Instruct-2507 via GRPO (400 rollouts of 25 environments), SPADE lifted the 30B-A3B game-environment suite average to 58.3, +8.1 over base, and improved tool-use results across backbones. The issue also references Hawkeye for building better GPU kernels.

Import AI·Aug 24, 13:12 UTC · 23d agoAI research

Atria Dawn: The Dawn of Agentic Superintelligence

Atria Dawn Preview, an agentic foundation model trained on verifiable experiences, tops five of 16 research and engineering benchmarks.

Atria Dawn Preview is a foundation agentic language model for scientific research and engineering workflows, trained via a Verifiable Experience Pipeline connecting tool-mediated interactions to executable environments and externally verified outcomes. Across 16 benchmarks spanning research, engineering, and digital work it is competitive with frontier agents and achieves the highest reported score on five of them. The release includes a human-AI collaboration case study analyzing 769 task records from 56 participants, where about one-third of completed AI-assisted tasks were rated infeasible without AI and agents frequently proposed methods and implemented revisions while humans retained final decisions.

Hugging Face daily papers·Sep 13, 20:00 UTC · 2d agoModel release

Engineered Persuasion: Evaluating Personalized Pretexts in LLM-Generated Spear Phishing

A study of 180 US workers found each LLM phishing personalization level raised click-intention odds by 28%, but credibility depends on context fit.

The arXiv paper evaluates how personalized pretexts in LLM-generated spear phishing affect perceived credibility, using 180 US working adults across 1,436 evaluations of emails with four cumulative personalization levels, from workplace context to shared-project details. Convincingness rose 2.40 points per level in sensitivity analysis and click-intention odds increased 28% per level, while non-clickers shifted toward deleting rather than reporting. Qualitative coding showed details matching the recipient's role and routines supported credibility, whereas incorrect, vague, or channel-inappropriate details raised suspicion. The authors argue personalization effectiveness depends on pretext fit, with implications for workplace security training.

arXiv cs.CR·Sep 3, 19:21 UTC · 12d agoResearch

Trusting-Trust Attack against an Entire Linux Distribution (via the strip utility)

ArXiv paper shows the trusting-trust compiler backdoor technique can compromise an entire Linux distribution via the strip utility.

The paper (arXiv 2607.24888) demonstrates that Ken Thompson's trusting-trust attack, long viewed as a compiler-specific threat, can backdoor an entire Linux distribution by targeting the strip utility. A compromised tool reproduces its backdoor in subsequent rebuilds of itself, generalizing the attack surface beyond compilers. The finding has supply-chain implications for build reproducibility and distribution trust, though it is a research result with no observed real-world exploitation.

Lobsters · security·Sep 5, 10:58 UTC · 11d agoResearch

What Good Identity Hardening Looks Like

Huntress guidance defines mature identity hardening beyond baseline MFA, covering exception cleanup and detecting identity configuration drift before attackers do.

Huntress argues MFA is only a starting point and outlines what mature identity hardening actually looks like in practice. The guidance covers closing MFA exceptions and coverage gaps that create unauthenticated attack paths. It also stresses catching identity configuration drift before attackers can exploit it.

Huntress·Aug 25, 13:00 UTC · 22d agoResearch

ENEAS: Embedding-guided Neural Ensemble for Adaptive Segmentation

ENEAS adds text prompting and semantic verification to video segmentation to keep tracking targets through occlusion and reject lookalike distractors.

ENEAS is a unified text-promptable method for instance tracking and open-concept semantic discovery in video, designed to fix temporal hallucinations, spatial fragmentation, and semantic misclassification seen in SAM 3-class foundation models. It extends the geometrically robust SeC architecture with a text-prompting adapter and temporal memory, and uses a verification layer combining fast visual embedding matching with conditional VLM refinement for ambiguous candidates. It targets 3D reconstruction pipelines where a single misclassified distractor corrupts the asset. Code and models are open-sourced.

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

AuK Technical Report: An Open-Source Foundational Model for Speech Generation and Editing

Open-source speech foundation model AuK unifies generation and editing, trained on 1.95 million hours, with distilled AuK-Flash achieving 4.5x speedup.

AuK is an open-source foundational model that unifies speech generation and editing through natural-language instructions and audio context, trained on approximately 3.03 billion instruction-audio instances and 1.95 million hours of supervision across five task families including generation, content editing, and acoustic editing. It combines a multimodal LLM for semantic conditioning, a VAE jointly trained on speech, general audio, and music, and a hybrid rectified-flow Transformer using dual-stream MMDiT blocks followed by unified single-stream DiT blocks. Post-training applies human-feedback preference optimization for editing and reward-based reinforcement learning for generation, and the distilled AuK-Flash performs 4-step inference without classifier-free guidance at a 4.5x wall-clock speedup. Source code and model weights are released.

Hugging Face daily papers·Sep 7, 20:00 UTC · 8d agoModel release2

Critical Unauthenticated File Upload to RCE in Elementor Pro Plugin

Elementor Pro WordPress plugin has an unauthenticated arbitrary file upload flaw in its Forms module allowing remote code execution.

Patchstack disclosed a critical unauthenticated arbitrary file upload vulnerability in the Elementor Pro WordPress plugin that can lead to remote code execution. The flaw is in the Forms module's File Upload field, where the extension check and the file-move step run in two separate loops with inconsistent handling of empty file entries. The writeup describes exploitation via crafted upload submissions; no CVE id or evidence of in-the-wild exploitation is given in the post.

Patchstack·Aug 19, 17:17 UTC · 28d agoVulnerability

Rosetta at AlexandriaX-2026: LoRA-Adapted NileChat for Context-Aware Dialectal Arabic Dialogue Translation

Rosetta ranks 4th and 5th in AlexandriaX-2026 dialectal Arabic dialogue translation using a LoRA adapter on NileChat-3B, finding limited pretraining benefit.

The Rosetta system for the AlexandriaX-2026 shared task fine-tunes a LoRA adapter on NileChat-3B for context-aware English-to-dialectal Arabic dialogue translation. The adapter was additionally pretrained on MADAR and PADIC dialect corpora for the unconstrained track. It achieved spBLEU 26.10 (4th, constrained) and 25.09 (5th, unconstrained). External dialect pretraining improved only two of thirteen dialects while slightly degrading overall performance, indicating negative transfer.

arXiv cs.AI / cs.LG / cs.CL·Sep 9, 16:15 UTC · 7d agoAI research1

ActReview: Rebuttal-Guided Training Data and Rubric Rewards for Actionable Peer Review Generation

ActReview post-trains Qwen3-8B-Base on 40K rebuttal-derived instances with rubric rewards to generate actionable, grounded peer-review feedback, plus a 1,000-instance benchmark.

The framework builds ActReview-40K from real OpenReview review-rebuttal threads, aligning reviewer weaknesses with author responses and grounding feedback in localized paper evidence. Qwen3-8B-Base is post-trained with multi-task supervised fine-tuning followed by GRPO using candidate-aware, weakness-specific rubric rewards. Experiments show improved actionability and grounding over prior specialized review-generation models, supported by ActReview-Bench, a human-curated 1,000-instance evaluation set. Human evaluation confirms better revision usefulness while noting a remaining gap in technical accuracy.

arXiv cs.AI / cs.LG / cs.CL·Sep 8, 17:30 UTC · 8d agoAI research

EVOHARNESSBENCH: Can Your Agents Keep Pace with an Evolving Harness?

Researchers introduce EVOHARNESSBENCH, a benchmark showing that evolving agent harnesses (tools, skills, agents) cause forgetting and inconsistent adaptation across 802 tasks.

The paper introduces EVOHARNESSBENCH, a benchmark that places non-stationarity in the externally supplied agent harness rather than in the task stream, evaluating agents across tools, skills, and specialist agents. It comprises 17 multi-stage harness streams built deterministically from verifier-based benchmarks, totaling 802 tasks, 520 tools, 42 skills, and 62 agents. Evaluation covers deployment (retention of previously accessible competence) and self-evolving adaptation settings. Results show harness expansion alone degrades previously solved tasks (harness-induced forgetting), adaptation gains are inconsistent, and retention and adaptation can pull in opposite directions.

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

[AINews] AEF-1 standard emerges for Third Party Evaluators, as Xai, OpenAI, and Anthropic all cosign

xAI, OpenAI, and Anthropic cosign the AEF-1 third-party evaluation standard while Dario Amodei proposes embedded evaluators for safety verification.

The AI Evaluator Forum published AEF-1, a baseline standard for independent third-party AI evaluations covering access, conflicts of interest, funding relationships, recusal, and transparency, cosigned by xAI, OpenAI, and Anthropic. Dario Amodei wrote a rare personal blogpost proposing embedded evaluators such as METR with desks, badges, company laptops, and internal-risk-team-level access to verify safety commitments, plus democratic and global coordination frameworks. The roundup also covers the pacing debate: Bilal Chughtai left Google DeepMind arguing progress may outrun alignment, while critics including Aidan Gomez and Cohere push back against slowdowns and lab gatekeeping. Additional items include Cline Desktop's launch with open-weight model support.

Latent Space·Sep 15, 04:50 UTC · 1d agoAI safety & security

Elementor Pro RCE Flaw Under Active Attack

A critical Elementor Pro Forms module flaw allowing unauthenticated file uploads is under active attack against widely used WordPress sites.

A critical vulnerability in Elementor Pro, a widely used WordPress page builder plugin, allowed unauthenticated attackers to upload arbitrary files through the plugin's Forms module. The SOCRadar report indicates the flaw is being actively exploited in the wild. No CVE identifier was provided in the available text.

SOCRadar·Sep 4, 13:34 UTC · 12d agoExploit / PoC in the wild

Multi-Vector (Late Interaction) Embedding Models with Sentence Transformers

Hugging Face details building and using multi-vector late-interaction embedding models with Sentence Transformers for retrieval workloads.

Hugging Face published a guide on multi-vector, late-interaction embedding models (ColBERT-style) supported through Sentence Transformers. The post covers how practitioners can build and use these models for retrieval and RAG pipelines. It is a developer tooling and technique write-up, not a security advisory.

Hugging Face Blog·Aug 18, 00:00 UTC · 29d agoAI tools & infra1

One Symptom, Three Levers: A Critical Review of On-Policy Self-Distillation

A review paper frames on-policy self-distillation collapse as governed by three levers: token weighting, privileged information, and guidance decay.

The paper critically reviews On-Policy Self-Distillation (OPSD), where a language model trains on its own generations scored token-by-token by a teacher conditioned on privileged information such as reference solutions or environment feedback. It identifies collapse, the progressive narrowing of producible reasoning paths, as the dominant failure mode and analyzes it through three levers: signal weighting, the nature of privileged information, and teacher dynamics. The review is restricted to mathematical reasoning, reports no new experiments, and offers a shared vocabulary separating settled findings from disputed ones.

Hugging Face daily papers·Aug 25, 20:00 UTC · 21d agoAI research

Post-DEF CON Phishing Uses Malicious Google Doc to Deliver Malware

Huntress uncovered post-DEF CON phishing via X direct messages using a malicious Google Doc to deliver AMOS and NetSupport RAT malware.

Huntress uncovered a phishing campaign targeting attendees after Black Hat and DEF CON. Attackers used X direct messages pointing to a malicious Google Doc as the delivery vehicle. Payloads include AMOS, a macOS infostealer, and the NetSupport RAT, among other malware.

Huntress·Aug 19, 13:00 UTC · 28d agoPhishing & fraud

HazardAuditor: From Executable Threats to Safer Computer-Use Agents

HazardAuditor trains execution-grounded guard models for computer-use agents, improving safety verdict accuracy by up to 16.5 points.

HazardAuditor runs heterogeneous agents (Claude Code, Codex, Hermes, OpenClaw) in controlled environments and normalizes their interactions into a canonical event representation for cross-framework supervision. It introduces Guard Policy Optimization (GuardPO), which converts deterministic safety outcomes into sequence-level advantages and normalizes rationale and verdict regions so the safety decision becomes the effective optimization unit. Across multiple benchmarks and heterogeneous computer-use systems, HazardAuditor improves accuracy by up to 16.5 percentage points over the strongest prior guard model. Code, models, and evaluation artifacts are being released.

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

The Frontier AEO Tracker: What Astra Chooses (and every other frontier model, and what you can do about it)

Latent Space launches an AEO tracker scoring 7 frontier models' product recommendations across 161 categories, revealing generational bias flips.

Latent Space built a tracker measuring Answer Engine Optimization by running 6 prompt variations across 7 frontier models with search enabled over 161 product categories, scoring first choices, alternatives, mentions, and anti-recommendations. It found 28 categories with a universally dominant primary choice and observed soft biases, such as models favoring their own lab's coding agents. Analysis of Anthropic's Sol→Astra and Opus→Fable generations showed newer models consulting fewer sources and being less likely to change answers when questions are paraphrased.

Latent Space·Sep 7, 21:32 UTC · 8d agoAI research