50
You Can't Prefer Emotions You Don't Sample: Intensity Undershoot in DPO-Tuned LLMs
Study quantifies DPO-tuned LLMs undershooting requested emotional intensity, tracing the gap to candidate-pool extremity rather than conditioning format.
Conditioning an instruction-tuned LLM on continuous valence-arousal targets yields gain of only 0.26 for valence and 0.13 for arousal on Llama-3.1-8B, far below faithful control of 1.0. The authors attribute undershoot to neutral-heavy preference corpora like EmoBank and candidate pools lacking extreme affect, leaving DPO without extreme exemplars. Uniform target coverage with a hotter candidate pool raises valence gain to 0.40 on Llama-3.1-8B and 0.44 on Qwen3-8B, with modest in-distribution cost; arousal gains remain unstable across seeds.
31
PLC-DPO: Posterior Label Correction in Noisy and Ambiguous Preference Optimization
PLC-DPO corrects noisy preference labels during DPO by routing each pair as clean, flip, or tie using calibrated policy-reference margins.
PLC-DPO addresses reversed, weak, or ambiguous preference labels in Direct Preference Optimization by using the calibrated policy-reference margin as online evidence to route each pair's training signal as clean, flip, or tie. This actively corrects supervision direction and strength rather than only filtering suspicious examples. Across 57 dataset-model-benchmark cells, PLC-DPO achieves the best mean win rate of 60.5 versus 55.5 for the next-best method. Injected-noise and tie stress tests plus human disagreement analysis show the routing stays stable and distinguishes flipped from weakly directional pairs.
28
Automatic Key Exchange: faster, post-quantum secure origin handshakes for 45 billion daily connections (and counting)
Cloudflare's Automatic Key Exchange probes origins to lead with preferred key exchange, cutting HelloRetryRequests from 52% to 3.7% and enabling automatic post-quantum origin connections.
Cloudflare announced Automatic Key Exchange, an extension of Automatic SSL/TLS that probes each origin's supported key agreement algorithms and leads with the preferred one, favoring the post-quantum hybrid X25519MLKEM768. Rollover across roughly 45 billion daily origin connections cut HelloRetryRequests from about 52% to 3.7%, saving more than 150 ms of p90 handshake latency. Hundreds of thousands of domains now receive post-quantum origin connections without manual configuration, supporting Cloudflare's goal of quantum-safe encryption by 2029 to counter harvest-now-decrypt-later attacks.
45
StepAudio 3 Music Technical Report
StepAudio 3 Music introduces long-form text-controlled music generation using ABC-notation planning and flow-matching diffusion, ranking near the top music arena.
StepAudio 3 Music generates long-form, text-controlled music using a 50-Hz single-codebook tokenizer with 65,536 entries and a flow-matching diffusion Transformer over VAE latents. A Mixture-of-Experts autoregressive model first plans an arrangement in ABC notation (ABC-CoT) before predicting music tokens. With DPO fine-tuning, it tops AudioBox content and production quality scores and reaches Quality Elo 1105 on the Artificial Analysis Music Arena, behind Suno V5.5 and Mureka. Generation covers songs, accompaniment from dry vocals, and cover synthesis up to 5 minutes 30 seconds at 48-kHz output.
30
5 useful things you'll learn in my new post-training textbook (shipping now!)
Nathan Lambert's new RLHF and post-training LLM textbook covers PPO, GRPO, GSPO, CISPO and related techniques, freely available online.
Nathan Lambert's book 'Reinforcement Learning from Human Feedback: Aligning and Post-training LLMs' is now shipping from Manning. It covers policy-gradient algorithms including PPO, GRPO, GSPO, CISPO, and RLOO, plus loss aggregation, truncated importance sampling, asynchronous RL systems, and post-training topics like rejection sampling, outcome reward models, and on-policy distillation. The book is freely available online with a 12-hour course, codebase, and exercises.
28
30
Jev: New frontier model 40-400x cheaper and 20-200x faster
TypeSafe AI launches Jev, an early-access 'System One' model delivering calibrated structured outputs claimed 40-400x faster and cheaper than LLMs.
TypeSafe AI, founded by former OpenAI researcher Diogo Almeida, released its first 'System One Model' called Jev in early access. Jev forgoes string generation and is trained with Reinforcement Learning for Calibrated Decisions (RLCD) to produce type-safe structured values with calibrated probabilities. The company claims 70-500ms response times (40-200x faster), input pricing of $0.042 per million tokens, and free output tokens via a parallel sampling architecture. Target use cases include AI-powered workflows, real-time applications, and verification/guardrail tasks.
45
RegionFed: Federated Learning for Personalized Query Understanding in Heterogeneous Retail Environments
RegionFed is a gradient-level federated learning framework enabling personalized retail query understanding while matching centralized accuracy with differential privacy.
RegionFed is an architecture-robust federated learning framework for personalized query understanding that operates at the gradient level, using the l2 conflict between regional and global gradients to diagnose heterogeneity and control personalization. Existing parameter-level personalized FL methods collapse on transformers, falling below 10% accuracy on T5, while RegionFed deploys unchanged on T5-Small, T5-3B, RoBERTa, and CNNs. RegionFed-Meta achieves 92.27% across Amazon ESCI, Amazon Reviews, and LEAF-FEMNIST, within 0.23 percentage points of the centralized upper bound, with epsilon-approx-0.60 differential privacy.
30
How Fyxer built an AI executive assistant people trust
Fyxer details its OpenAI-powered AI executive assistant, orchestrating 30-50 specialized models trained on 500,000+ hours of assistant workflows.
OpenAI published a case study on Fyxer, whose AI executive assistant orchestrates 30-50 specialized OpenAI models trained on more than 500,000 hours of annotated executive assistant workflows. The system uses supervised fine-tuning, LoRA, and Direct Preference Optimization on user edits, and 53% of AI-generated email drafts are accepted as written. Fyxer's annual recurring revenue grew from $1 million to $32 million during 2025.
15
Mind2Dialogue: Training Human-Aware Language Models by Simulating User Mental States
Mind2Dialogue simulates users' mental states to generate privileged supervision, boosting personalization and preference-following in Qwen, Llama, and OLMo assistants.
The Mind2Dialogue framework uses a psychology-guided simulator that preserves personal characteristics while updating user mental states through interaction, driving coherent conversations and an Oracle assistant's responses. Privileged distillation trains models on the Oracle's well-informed responses so they can assist users without direct access to mental states at deployment. Training on the full corpus improves every reported personalization metric over Qwen, Llama, and OLMo instruction-tuned baselines, including 26.6 to 40.9 percentage point gains in preference-following generation.
45
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.
42
Meta FAIR Introduces AI Research Preference Models (RPMs): Ranking ML Experiments Before Spending GPU Hours
Meta FAIR, Oxford and UCL introduce Research Preference Models that rank unexecuted ML experiments, lifting AIRS-Bench scores from 0.684 to 0.729 and cutting compute ~1.6×.
Researchers from Meta FAIR, Oxford, and UCL introduce Research Preference Models (RPMs), which use frozen pretrained LLMs (Qwen3.6-27B backbone, no fine-tuning) to rank unexecuted experiment candidates and execute only the winner of a pairwise knockout tournament. Two variants shipped: an inference-only LLM-as-a-judge and an agentic variant that runs small pilot experiments in an H200 sandbox. On AIRS-Bench (20 tasks, 24 hours on one H200, 10 seeds), scores rise from 0.684 (random) to 0.711 and 0.729 versus a 0.748 validation oracle, and both variants reach the baseline's 24-hour score in roughly 15 hours. The team reports new SOTA on WinoGrande (94.1% with Agentic RPM) and SVAMP (95.7% with inference-only).
45
Domain-Specific Hallucination Detection in Large Language Models
A multi-signal pipeline detects LLM hallucinations, reaching F1 0.915 on HaluEval and cutting Qwen2.5-0.5B hallucination rates from 85.5% to 37.7% via DPO.
The paper presents a hallucination detection pipeline combining fine-tuned DeBERTa-v3 classification, Monte Carlo Dropout uncertainty, and temperature-scaled calibration. It achieves F1 0.915 and AUROC 0.977 on general-domain HaluEval tasks, with MC Dropout inference raising accuracy to 93.2%. Applying DPO to a Qwen2.5-0.5B generator reduces its hallucination rate from 85.5% to 37.7%, while cross-domain evaluation shows poor general-domain transfer to SciFact (F1 0.52) and PubMedBERT fine-tuning as the strongest adaptation (F1 0.63).
32
50
X's Algorithm Feeds Off Ragebait and Impacts Democrats More, Study Finds
A PNAS study of 715 X users finds the platform's engagement-optimizing algorithm amplifies value-misaligned ragebait, affecting self-identified Democrats more.
A study published in the Proceedings of the National Academy of Sciences used browser extension data from 715 U.S. X users, recruited in September and October 2024, to compare self-reported values with the content the For You feed amplified. It found that replying to posts—only 6.8% of observed interactions—disproportionately shapes the engagement-optimizing algorithm, creating runaway feedback loops of value-misaligned ragebait that were more pronounced for self-identified Democrats. Co-author Ziv Epstein, a postdoctoral researcher at Stanford University, said the observational work is intended to spur debate on algorithm transparency and user control over feeds.
15
30
GPT-6 Astra: an automated AI Engineer you can hire for <$6 an hour
OpenAI launches GPT-6 Astra, a frontier model scoring 97.6% on FrontierMath and 99.9% on ARC-AGI-3, capable of autonomous AI engineering at roughly $6 per hour.
OpenAI launched GPT-6 Astra, described as its first Stargate and lightly looped frontier model, beating Fable 5.1 on many metrics and saturating the hardest FrontierMath (97.6%) and ARC-AGI-3 (99.9%) benchmarks. Latent Space tested the model with over 20 billion tokens, reporting it can train and select models, label data, deploy and debug systems, and orchestrate 20-50 parallel subagents. The authors measured about $6 per hour of agentic engineering at 33 tokens per second, with token efficiency independently confirmed by Artificial Analysis.
93
30
AgenticGen: Reward-Guided Agentic Video Generation for Advertising
AgenticGen applies DPO and GRPO reward-guided reasoning to ad video generation, improving TikTok CTR 2.72%, CVR 2.63%, and Advv 9.61%.
AgenticGen decomposes advertising video generation into two trainable reasoning stages, strategy selection and draft generation, supervised by online business feedback. It learns a performance-based reward from accumulated online feedback plus a rubric-based reward aligned with human quality standards, then optimizes policies with DPO followed by GRPO using process and outcome rewards. Online A/B experiments in the TikTok advertising system show CTR up 2.72%, CVR up 2.63%, and Advv up 9.61% over an SFT baseline.
26
30
[AINews] 10% worse, 100x cheaper, 10000x faster: Why Simulation is taking over
Latent Space argues AI training pipeline stages—rewards, data, teachers, curricula, environments—are flipping from human-made to model-made simulation.
Latent Space's AINews essay traces how each component of AI training has turned synthetic since 2022: reward models (InstructGPT, RLAIF), synthetic pretraining data (Microsoft Phi, NVIDIA Nemotron-4 340B), model teachers (Alpaca, DeepSeek-R1 distillation), and self-generated curricula (Self-Rewarding Language Models, SPIN). In 2026 it highlights Karpathy's autoresearch loop—700 experiments yielding 20 kept improvements, cutting GPT-2 training time from 2.02 to 1.80 hours—and Z.ai's GLM-5.3 fully synthetic RL environment, judging, and verification stack. It frames these shifts as 'simulation': 10% worse but 100x cheaper and 10,000x faster than human equivalents.
35
AI for everyone in every language
Google says its AI now spans 300+ languages reaching 7 billion people, unveiling Gemini 3.5 Transcribe, Live Translate, and TranslateGemma models.
Google announced its technologies now support more than 300 languages spoken by 7 billion people, 86% of the global population. Gemini 3.5 Live Translate powers real-time spoken translation across 70 languages and 2,000+ language pairs, while Gemini 3.5 Transcribe is its most precise speech-to-text model. Its Universal Speech Model was trained on 12 million hours of audio using cross-lingual transfer learning, and TranslateGemma is a family of lightweight open translation models covering 55 languages that run on-device. Open-data partnerships include WAXAL covering 27 Sub-Saharan African languages and Project Vaani with 30,000+ hours of speech across 109 languages.
58
30
Opaque recurrence, and other AI terms that you should probably know
TechCrunch updates its plain-English glossary defining common AI terms from AGI and agents to chain-of-thought reasoning.
TechCrunch maintains a regularly updated glossary of AI terminology, defining terms such as AGI, AI agents, API endpoints, chain of thought, coding agents, compute, deep learning, and diffusion. It highlights 'opaque recurrence', the reasoning technique in OpenAI's new Astra model that has drawn attention from AI safety researchers. The piece is an educational living document rather than new research or a product announcement.
18
30
FlowBalance: Verifier-Grounded Self-Improvement from On-Policy Reasoning Experience
FlowBalance is a verifier-grounded self-improvement method that beats FlowRL on Qwen3-4B and Qwen3-8B math reasoning while improving training stability.
FlowBalance calibrates dense self-guidance scores with verifier-derived group advantages: guidance is retained on positive-advantage trajectories, reversed on negative-advantage trajectories, and disabled when rollout groups show no outcome preference. The method exponentially reweights a reference policy via trajectory balance, with guarantees including within-group contrast preservation and a minimum-change reverse-KL characterization. On mathematical reasoning it outperforms FlowRL on Qwen3-4B and Qwen3-8B, trains faster and more stably, avoids direct OPSD's response-length collapse, and shows higher correct-strategy diversity on AIME24.
32
PhysStream: Streaming Physics-Grounded Video Generation with Structured Scene Memory and Fine-Grained Motion Control
PhysStream enables mid-generation interactive control of physics-grounded video via structured scene memory and velocity-increment signals, reducing motion distribution distance 33%.
PhysStream is an autoregressive physics-grounded image-to-video model that maintains structured scene memory—positional maps and object tracking maps derived online from previously generated frames—and accepts fine-grained motion control via sparse velocity-increment signals encoding physical quantities. Training runs in two stages: a bidirectional model finetuned with motion-control conditioning, then a causal autoregressive model with structured scene memory. It supports interactive mid-generation control over multi-object tabletop rigid-body scenes, reducing motion distribution distance (FVMD) by 33% and trajectory error by 12% over the strongest baselines. Human evaluators preferred it in over 85% of in-the-wild comparisons.
28
30
Tell HN: OpenAI brings back 5 hour limit for plus and business standard users
OpenAI reinstated a 5-hour usage limit for Plus and Business Standard subscribers, sparking Hacker News debate about subsidized inference pricing and subscription value.
A Hacker News discussion reports that OpenAI has brought back a 5-hour usage limit for Plus and Business Standard users. Commenters debated whether cheap tokens are a subsidized customer acquisition strategy, whether AI companies have sustainable margins on inference, and how switching costs between providers affect dependency. The thread reflects community sentiment on pricing and usage caps rather than an official policy announcement.
35
Cohesity adds recovery capabilities for AI agents and the data they manage
Cohesity launched Agent Resilience to discover, protect, and recover AI agent memory, configuration, and agent-managed data, debuting with Amazon Bedrock integration.
At Cohesity Catalyst, Cohesity introduced Agent Resilience within Cohesity Data Cloud, protecting AI agent memory and configuration with snapshot architecture, immutable backups, and clean-room recovery, plus recovery for databases and file systems that agents manage. It launches with Amazon Bedrock integration, support for Microsoft and Google platforms planned, and general availability targeted for year-end. The company cited Gartner's prediction that up to 40% of enterprise applications will include task-specific agents by 2026, and Cohesity research showing 56% of organizations are unprepared to detect or contain unintended agent actions while 58% lack confidence in verifying AI model integrity after attacks. Cohesity also outlined an Autonomous Cyber Resilience vision using agentic workflows and introduced the AI Resilience Academy.
30
ChatGPT can now connect to your personal apps to mimic writing style
OpenAI is testing ChatGPT Writing Style, which mimics a user's voice using writing samples from connected Gmail, Slack, and Drive accounts.
OpenAI confirmed it is testing a Writing Style feature for ChatGPT that learns a user's voice from writing samples in connected apps. The onboarding flow references Messaging (Slack), Documents (Google Drive and Notion), and Email (Gmail) as example sources. It resembles Anthropic's Styles personalization feature but draws on existing writing inside connected services rather than uploaded samples. The feature is available to a small group of users with no announced general rollout date.
28
30
CISA's logging guidance works beyond government
CISA released its Logging Reference Architecture in August 2026 to help federal agencies meet OMB M-26-14 logging requirements, usable as a benchmark by critical infrastructure operators.
CISA's Logging Reference Architecture (LRA), released in August 2026, helps US federal civilian agencies satisfy logging requirements in OMB Memorandum M-26-14 and explicitly encourages critical infrastructure operators to use it as a benchmark. The framework is organized around continuous event monitoring and threat hunting, investigation, response, and forensics, with a federal baseline of six months searchable and one year retrievable logs. Agencies must submit Agency Logging Plans within 90 days and work toward Advanced maturity within 320 days; the guidance also treats AI outputs as derived data requiring human review and preserved metadata.
34
Instacart launches an AI grocery shopping assistant called Clementine
Instacart launches Clementine, a conversational AI grocery assistant that turns chats, lists, or recipes into purchasable carts in the US and Canada.
Instacart launched Clementine, an AI assistant that converts conversations, grocery lists, photos of handwritten lists, or recipes into ready-to-buy carts, now available in the US and Canada. It generates personalized recipes, tailors recommendations to dietary needs such as gluten-free and vegetarian, and surfaces deals and lower-cost alternatives. Uber Eats and DoorDash have launched similar AI assistants as delivery apps compete to become households' default meal-planning tools.
15
30
30
Can your coding style predict whether your code is vulnerable?
University of Massachusetts Dartmouth researchers present VulStyle, a stylometry-based vulnerability detector that also exposes benchmark reliability problems.
VulStyle combines stylometric features with syntax-tree structure and source tokens, pre-trained on about 4.9 million functions across seven programming languages and fine-tuned on five vulnerability detection datasets. It beat token-only detectors on some benchmarks but its F1 drops sharply on DiverseVul, which the authors link to noisy labels inflating reported performance across popular datasets. The authors argue style-aware detection should be harder to evade but did not test this empirically, and they note that uniform LLM-generated code may strip away the individual developer style the model depends on.
48
30
Explainability Assistant: A Conversational XAI Interface for Interpreting Energy Consumption Models
Researchers release Explainability Assistant, an open-source conversational XAI tool using LLM function calling, lifting intent-parsing accuracy from 76.8% to 94%.
The paper introduces the Explainability Assistant, an open-source conversational XAI system for interpreting energy consumption forecasting models such as genetic-programming symbolic regressors. It uses LLM function calling instead of rigid custom grammars, achieving 94% intent-parsing accuracy versus 76.8% for prior work TalkToModel, and adapts to different ML problem types without task-specific fine-tuning. Comparative evaluation with energy domain specialists against a traditional XAI dashboard showed improved usability, with all experts preferring the conversational interface.
25