30
30
50
30
50
30
30
30
30
30
30
30
35
Coupled Calibration and Learning: Mitigating Teacher Bias in LLM Distillation without Target-Domain Reward Feedback
CCL couples teacher calibration with student updates via token-level branching, provably removing teacher bias in LLM distillation.
The paper proposes Coupled Calibration and Learning (CCL), an LLM distillation algorithm that alternates teacher calibration using source-question reward feedback with student training on target questions under covariate shift. Each iteration calibrates the teacher on source feedback, trains the student on target questions, and lets the updated student inform subsequent calibration. The authors prove the student's expected KL divergence to the oracle student converges to zero at a polynomial rate, and show regularized direct matching error can remain bounded away from zero.
20
30
30
30
30
30
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
Beyond Solver Verdicts: Generative Reward Models for Autoformalization
Researchers introduce Generative Verification (GenV), a generative reward model achieving 0.961 AUROC in detecting unfaithful autoformalization that preserves solver verdicts.
The paper formalizes Verdict-Preserving-Unfaithfulness (VPU), a failure mode in neurosymbolic autoformalization where an incorrect encoding executes successfully and matches the expected solver verdict, and proves verdict-only verification is bounded to chance-level detection. The proposed Generative Verification (GenV) distills an offline Z3-equivalence oracle into a reference-free, continuous reference-equivalence score within the language model's vocabulary space. The oracle-mined verifier (GenV+HN) achieves 0.961 AUROC, generalizes zero-shot across unseen translators and formal styles, and yields an 11.3-point downstream accuracy gain in agentic test-time compute allocation. Mechanistic analysis with decision-projected logit lenses and sparse autoencoders shows the generative readout extracts precise spatial error coordinates without explicit localization training.
35
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
30
30
30
⚡ Weekly Recap: Telecom Sleeper Cells, LLM Jailbreaks, Apple Forces U.K. Age Checks and More
50
30
30
30
30
35
30
NVIDIA Vera Rubin NVL72 Delivers Leading Performance in MLPerf Inference v6.1 Debut
NVIDIA's Vera Rubin NVL72 debuts in MLPerf Inference v6.1 with up to 3.7x higher throughput than GB300 NVL72 and 99% scaling efficiency at 288 GPUs.
In its first MLPerf Inference preview submission, NVIDIA's Vera Rubin NVL72 achieved up to 3.7x higher throughput than GB300 NVL72 on Qwen3-VL and 2.5x on DeepSeek-R1. A 288-GPU GB300 NVL72 submission across four racks reached 99% scaling efficiency on the DeepSeek-R1 offline benchmark. Software optimizations delivered up to 1.6x gains over v6.0, leveraging TensorRT-LLM, vLLM, Dynamo, disaggregated serving, and NVFP4 precision.
42
What must happen for AI’s trillion-dollar gamble to pay off
Hyperscalers need 2.7x productivity gains by 2030 to justify nearly $1.1 trillion in AI data center spending, or risk bankruptcy and capital misallocation.
Wharton finance professor Jessica Wachter estimates hyperscaler AI expenditure will reach nearly $1.1 trillion through 2027 and that a 2.7x productivity increase is needed to break even by 2030. AI revenues of roughly $150-200 billion this year fall far short of about $750 billion in annual spending, with total investment from Alphabet, Microsoft, Amazon, Meta, and Oracle potentially exceeding $5 trillion over four years. Alphabet reported its first free cash flow deficit (about $5.9 billion) since its 2004 IPO due to AI infrastructure costs. Researchers warn that failed demand could make the buildout the largest capital misallocation in history, with depreciating GPU chips risking stranded assets.
35
ImpossibleRubrics: Stress-Testing Generated Rubrics as Reward Signals
ImpossibleRubrics benchmark shows LLM-generated rubric reward signals are exploited 8-26% of the time by adversarial answers on impossible tasks.
ImpossibleRubrics is a benchmark of 169 impossible tasks across six impossibility categories, each paired with a verifiable oracle certificate, plus 48 answerable controls, for stress-testing LLM-generated rubrics used as reward signals. Eleven rubric generators were exploited 8-26% of the time on an unbiased 150-task cut and up to 36% on a stress cut, while a certificate-faithful rubric scored 0%. A single generic 'be decisive, penalize hedging' rubric was exploited 64% of the time, suggesting tailored criteria can reveal which claims attackers should fabricate.
45
Pick Your Poison: Learning to Select Poison Sets for Stronger LLM Backdoor Attacks
Poison set selection swings LLM backdoor attack success from 3% to 80%; SAILS boosts held-out success by 30 points.
The paper shows that random poison set selection severely underestimates worst-case backdoor vulnerability: across three LLaMA-3-8B settings with fixed model, clean data, and poison count, attack success ranges from 3% to 80% depending only on which poison set is chosen. The authors formalize poison selection as oracle-budgeted set optimization and introduce SAILS, which learns a set scorer from a few hundred finetune-and-evaluate runs, ranks millions of candidate sets, and audits a shortlist. SAILS improves held-out attack success by 30 percentage points over the strongest influence baselines, transfers from small-scale to full-scale finetuning, and extends to code-generation, agentic, and API-only backdoors.
55
E2A-Bench: Benchmarking Evidence-to-Action Reliability in Financial Chart Reasoning
E2A-Bench, a 969-query financial chart reasoning benchmark, finds VLMs fail evidence-to-action consistency, with fine-tuning amplifying BUY:SELL bias 4-6x.
E2A-Bench is a 969-query benchmark built from 323 HS300 constituents across three input modalities with deterministic OHLCV-derived evidence anchors, evaluating grounding, reasoning-action consistency, evidence-confidence calibration, and directional coverage via UCR, RCI, ECI, and NDR metrics. Testing 20 VLMs showed the lowest-hallucination model ranked near the bottom on coverage with only 6.4% directional coverage, and oracle-aided verification reduced unsupported claims but could collapse coverage. Financial fine-tuning amplified the BUY:SELL ratio by factors of 4.21 to 4.68 across base-fine-tuned pairs.
34
OpenAI Launches the Agents API in Public Beta, Putting the Codex Harness Behind One API Call
OpenAI released its Agents API in public beta, exposing the managed Codex harness with hosted or self-hosted sandboxes, MCP tools, and subagents.
The Agents API is a managed service built on the open-source Codex harness, handling context compaction, tool search, programmatic tool calling, and multi-agent orchestration. Agents run in OpenAI-hosted sandboxes, self-hosted environments, or partner sandboxes from Blaxel, Cloudflare, Daytona, DigitalOcean, E2B, Modal, Oracle, Runloop, and Vercel. Data residency is US-only and Zero Data Retention is unsupported. Examples use model gpt-6-astra; vendor-reported results include SafetyKit cutting case review cost 60% and Ciridae achieving 4x lower subagent latency.
78
Now everyone can put data to work
OpenAI launched a Data agent in ChatGPT Work that connects to enterprise warehouses and builds shareable analysis dashboards without SQL.
OpenAI introduced a Data agent in ChatGPT Work that connects to approved sources including Snowflake, BigQuery, Databricks, Redshift, ClickHouse, MongoDB, and Datadog, plus files from Google Drive and SharePoint. It investigates metric changes, builds interactive dashboards, and integrates with BI tools such as Power BI, Tableau, Omni, Sigma, and ThoughtSpot using semantic layers from dbt, Databricks Genie Ontology, and Snowflake Horizon. Queries enforce the connected account's existing table, row, and column permissions, with administrators controlling access via Workspace settings. OpenAI says nearly all of its product team and over two-thirds of its GTM organization use it internally, and NTT Data, Thermo Fisher, and ServicePiston are Alpha customers.
62
IdeaAMBIG: Benchmarking Implementation-Critical Gaps in Research-Idea Specifications
IdeaAMBIG benchmark with 660 instances measures whether LLMs can spot and fix underspecified research-method details for faithful implementation.
Researchers introduce IdeaAMBIG, a benchmark of 660 evidence-grounded instances (163 real-world gaps from reproducibility reports and GitHub issues, 497 controlled synthetic gaps) built from papers, codebases, and reproduction artifacts. It evaluates codification-readiness assessment, defect localization, and clarification action generation. Across 13 LLMs, the best model achieved only 9.6% Macro Defect Recovery Rate on real-world instances but 80.6% clarification success when given the annotated defect. An oracle study showed gold resolutions raise the codification-ready rate from 14% to 98%, identifying defect localization as the main bottleneck.
28