home.social

#iclr — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #iclr, aggregated by home.social.

  1. Benjamin Han @[email protected] ·

    A multi-agent LLM where each agent learns when to defer to a human, trained with GRPO on a cost-aware reward. Each defer event becomes SFT data, so the model gradually absorbs the human's expertise. Tunable cost knob trades accuracy against human-call budget at deployment, no retraining.

    benjaminhan.net/posts/20260520

    #ICLR #HumanInTheLoop #AgenticSystems #Metacognition #RL #AI

    #iclr #humanintheloop #agenticsystems #metacognition #rl #ai
  2. Benjamin Han @[email protected] ·

    A multi-agent LLM where each agent learns when to defer to a human, trained with GRPO on a cost-aware reward. Each defer event becomes SFT data, so the model gradually absorbs the human's expertise. Tunable cost knob trades accuracy against human-call budget at deployment, no retraining.

    benjaminhan.net/posts/20260520

    #ICLR #HumanInTheLoop #AgenticSystems #Metacognition #RL #AI

    #iclr #humanintheloop #agenticsystems #metacognition #rl #ai
  3. Benjamin Han @[email protected] ·

    A multi-agent LLM where each agent learns when to defer to a human, trained with GRPO on a cost-aware reward. Each defer event becomes SFT data, so the model gradually absorbs the human's expertise. Tunable cost knob trades accuracy against human-call budget at deployment, no retraining.

    benjaminhan.net/posts/20260520

    #ICLR #HumanInTheLoop #AgenticSystems #Metacognition #RL #AI

    #iclr #humanintheloop #agenticsystems #metacognition #rl #ai
  4. Benjamin Han @[email protected] ·

    A multi-agent LLM where each agent learns when to defer to a human, trained with GRPO on a cost-aware reward. Each defer event becomes SFT data, so the model gradually absorbs the human's expertise. Tunable cost knob trades accuracy against human-call budget at deployment, no retraining.

    benjaminhan.net/posts/20260520

    #ICLR #HumanInTheLoop #AgenticSystems #Metacognition #RL #AI

    #ai #rl #metacognition #agenticsystems #humanintheloop #iclr
  5. Benjamin Han @[email protected] ·

    A multi-agent LLM where each agent learns when to defer to a human, trained with GRPO on a cost-aware reward. Each defer event becomes SFT data, so the model gradually absorbs the human's expertise. Tunable cost knob trades accuracy against human-call budget at deployment, no retraining.

    benjaminhan.net/posts/20260520

    #ICLR #HumanInTheLoop #AgenticSystems #Metacognition #RL #AI

    #iclr #humanintheloop #agenticsystems #metacognition #rl #ai
  6. Habr @[email protected] ·

    О чём говорили на ICLR 2026? Репортаж AIRI о поездке на конференцию в Рио

    Конференции в науке об ИИ очень любят и ждут. Подача работы на какое‑либо мероприятие из верхушки рейтинга CORE обычно престижнее, чем подача её же в журнал первого квартиля. В «большую тройку» главных конференций года принято включать NeurIPS, ICML и ICLR. Последняя обычно проходит раньше двух других — в этом году она прошла в конце апреля в Рио‑де‑Жанейро. Мы посетили ICLR 2026 вместе с коллегами из AIRI и рассказываем, чем запомнилась нашим исследователям эта командировка.

    habr.com/ru/companies/airi/art

    #iclr_2026 #iclr #искусственный_интеллект #машинное+обучение #конференция

    #конференция #машинное #искусственный_интеллект #iclr #iclr_2026
  7. Ireland @[email protected] ·

    europesays.com/ie/483750/ ‘Basically a Chinese racket’: how China streaked ahead at a global AI conference #AI #AIIndexReport #AIResearch #Alibaba #ArtificialIntelligence #ArtificialIntelligence #Bytedance #CarnegieMellonUniversity #China #ChineseUniversities #DmytroLopushanskyy #Éire #Google #HongKong #Huawei #ICLR #IE #InternationalConferenceOnLearningRepresentations #Ireland #MassachusettsInstituteOfTechnology #Meta #Technology

    #technology #meta #massachusettsinstituteoftechnology #ireland #internationalconferenceonlearningrepresentations #ie
  8. Benjamin Han @[email protected] ·

    SCoRe is a two-stage on-policy RL recipe that teaches a language model to revise its own answers using only self-generated data. On Gemini 1.5 Flash and 1.0 Pro it gains 15.6 points on MATH and 9.1 on HumanEval over the base model. At matched inference budgets, sequential self-correction beats parallel sampling up to 32 samples.

    benjaminhan.net/posts/20260512

    #Paper #LLMs #RL #Metacognition #Reasoning #ICLR #AI

    #paper #llms #rl #metacognition #reasoning #iclr
  9. Benjamin Han @[email protected] ·

    SCoRe is a two-stage on-policy RL recipe that teaches a language model to revise its own answers using only self-generated data. On Gemini 1.5 Flash and 1.0 Pro it gains 15.6 points on MATH and 9.1 on HumanEval over the base model. At matched inference budgets, sequential self-correction beats parallel sampling up to 32 samples.

    benjaminhan.net/posts/20260512

    #Paper #LLMs #RL #Metacognition #Reasoning #ICLR #AI

    #paper #llms #rl #metacognition #reasoning #iclr
  10. Benjamin Han @[email protected] ·

    SCoRe is a two-stage on-policy RL recipe that teaches a language model to revise its own answers using only self-generated data. On Gemini 1.5 Flash and 1.0 Pro it gains 15.6 points on MATH and 9.1 on HumanEval over the base model. At matched inference budgets, sequential self-correction beats parallel sampling up to 32 samples.

    benjaminhan.net/posts/20260512

    #Paper #LLMs #RL #Metacognition #Reasoning #ICLR #AI

    #paper #llms #rl #metacognition #reasoning #iclr
  11. Benjamin Han @[email protected] ·

    SCoRe is a two-stage on-policy RL recipe that teaches a language model to revise its own answers using only self-generated data. On Gemini 1.5 Flash and 1.0 Pro it gains 15.6 points on MATH and 9.1 on HumanEval over the base model. At matched inference budgets, sequential self-correction beats parallel sampling up to 32 samples.

    benjaminhan.net/posts/20260512

    #Paper #LLMs #RL #Metacognition #Reasoning #ICLR #AI

    #ai #iclr #reasoning #metacognition #rl #llms
  12. Benjamin Han @[email protected] ·

    SCoRe is a two-stage on-policy RL recipe that teaches a language model to revise its own answers using only self-generated data. On Gemini 1.5 Flash and 1.0 Pro it gains 15.6 points on MATH and 9.1 on HumanEval over the base model. At matched inference budgets, sequential self-correction beats parallel sampling up to 32 samples.

    benjaminhan.net/posts/20260512

    #Paper #LLMs #RL #Metacognition #Reasoning #ICLR #AI

    #paper #llms #rl #metacognition #reasoning #iclr
  13. Benjamin Han @[email protected] ·

    Let's Verify Step by Step compares process and outcome supervision on MATH. The process-reward model reaches 78.2% best-of-1860 vs 72.4% for outcome. But that gap narrows fast at small N, where most deployments actually live.

    benjaminhan.net/posts/20260512

    #Paper #LLMs #Reasoning #Mathematics #ICLR #OpenAI #AI

    #paper #llms #reasoning #mathematics #iclr #openai
  14. Benjamin Han @[email protected] ·

    Conformal Language Modeling (CLM) adapts conformal prediction to generative LMs: sample candidates, stop when a calibrated rule fires, return a set guaranteed to contain an acceptable answer. The more interesting half is the component-level filter — per-phrase coverage, not just set-level. That's the primitive for hallucination flagging: highlight the vetted phrases, leave the rest for review.

    benjaminhan.net/posts/20260505

    #ConformalPrediction #LLMs #Hallucination #ICLR #AI

    #conformalprediction #llms #hallucination #iclr #ai
  15. Benjamin Han @[email protected] ·

    Conformal Language Modeling (CLM) adapts conformal prediction to generative LMs: sample candidates, stop when a calibrated rule fires, return a set guaranteed to contain an acceptable answer. The more interesting half is the component-level filter — per-phrase coverage, not just set-level. That's the primitive for hallucination flagging: highlight the vetted phrases, leave the rest for review.

    benjaminhan.net/posts/20260505

    #ConformalPrediction #LLMs #Hallucination #ICLR #AI

    #conformalprediction #llms #hallucination #iclr #ai
  16. Benjamin Han @[email protected] ·

    Conformal Language Modeling (CLM) adapts conformal prediction to generative LMs: sample candidates, stop when a calibrated rule fires, return a set guaranteed to contain an acceptable answer. The more interesting half is the component-level filter — per-phrase coverage, not just set-level. That's the primitive for hallucination flagging: highlight the vetted phrases, leave the rest for review.

    benjaminhan.net/posts/20260505

    #ConformalPrediction #LLMs #Hallucination #ICLR #AI

    #conformalprediction #llms #hallucination #iclr #ai
  17. Benjamin Han @[email protected] ·

    Conformal Language Modeling (CLM) adapts conformal prediction to generative LMs: sample candidates, stop when a calibrated rule fires, return a set guaranteed to contain an acceptable answer. The more interesting half is the component-level filter — per-phrase coverage, not just set-level. That's the primitive for hallucination flagging: highlight the vetted phrases, leave the rest for review.

    benjaminhan.net/posts/20260505

    #ConformalPrediction #LLMs #Hallucination #ICLR #AI

    #ai #iclr #hallucination #llms #conformalprediction
  18. Benjamin Han @[email protected] ·

    Conformal Language Modeling (CLM) adapts conformal prediction to generative LMs: sample candidates, stop when a calibrated rule fires, return a set guaranteed to contain an acceptable answer. The more interesting half is the component-level filter — per-phrase coverage, not just set-level. That's the primitive for hallucination flagging: highlight the vetted phrases, leave the rest for review.

    benjaminhan.net/posts/20260505

    #ConformalPrediction #LLMs #Hallucination #ICLR #AI

    #conformalprediction #llms #hallucination #iclr #ai
  19. Benjamin Han @[email protected] ·

    MASS optimizes multi-agent LLM systems by interleaving prompt and topology search: block-level prompts, topology rejection sampling, then workflow-level prompts.

    Topology gets quietly demoted. Ablation on Gemini 1.5 Pro: ~6% gain from block prompts, 3% from topology, 2% from workflow prompts. Prompt tuning dominates — contradicts the topology-first thesis of ADAS and AFlow.

    benjaminhan.net/posts/20260430

    #LLMs #AI #AgenticSystems #PromptEngineering #Google #ICLR

    #llms #ai #agenticsystems #promptengineering #google #iclr
  20. Benjamin Han @[email protected] ·

    MASS optimizes multi-agent LLM systems by interleaving prompt and topology search: block-level prompts, topology rejection sampling, then workflow-level prompts.

    Topology gets quietly demoted. Ablation on Gemini 1.5 Pro: ~6% gain from block prompts, 3% from topology, 2% from workflow prompts. Prompt tuning dominates — contradicts the topology-first thesis of ADAS and AFlow.

    benjaminhan.net/posts/20260430

    #LLMs #AI #AgenticSystems #PromptEngineering #Google #ICLR

    #llms #ai #agenticsystems #promptengineering #google #iclr
  21. Benjamin Han @[email protected] ·

    MASS optimizes multi-agent LLM systems by interleaving prompt and topology search: block-level prompts, topology rejection sampling, then workflow-level prompts.

    Topology gets quietly demoted. Ablation on Gemini 1.5 Pro: ~6% gain from block prompts, 3% from topology, 2% from workflow prompts. Prompt tuning dominates — contradicts the topology-first thesis of ADAS and AFlow.

    benjaminhan.net/posts/20260430

    #LLMs #AI #AgenticSystems #PromptEngineering #Google #ICLR

    #llms #ai #agenticsystems #promptengineering #google #iclr
  22. Benjamin Han @[email protected] ·

    MASS optimizes multi-agent LLM systems by interleaving prompt and topology search: block-level prompts, topology rejection sampling, then workflow-level prompts.

    Topology gets quietly demoted. Ablation on Gemini 1.5 Pro: ~6% gain from block prompts, 3% from topology, 2% from workflow prompts. Prompt tuning dominates — contradicts the topology-first thesis of ADAS and AFlow.

    benjaminhan.net/posts/20260430

    #LLMs #AI #AgenticSystems #PromptEngineering #Google #ICLR

    #iclr #google #promptengineering #agenticsystems #ai #llms
  23. Benjamin Han @[email protected] ·

    DSPy turns LM pipelines into typed-module graphs and compiles them end-to-end against a single metric, bootstrapping its own few-shot demonstrations.

    The programming-model layer is the real contribution, not any specific teleprompter. Once pipelines are typed graphs, pipeline-level search (MASS, MIPRO) becomes possible in a way it wasn't with string-template prompts.

    benjaminhan.net/posts/20260430

    #LLMs #AI #PromptEngineering #NLP #Stanford #ICLR

    #llms #ai #promptengineering #nlp #stanford #iclr
  24. Benjamin Han @[email protected] ·

    EvoPrompt runs an evolutionary search over a population of prompts, with an LLM implementing crossover and mutation. Differential Evolution beats Genetic Algorithm on most BIG-Bench Hard tasks.

    One of the cleanest early examples of an LLM as *operator* in an optimization loop, not as the thing being optimized. That pattern then shows up across prompt-and-agent design: DSPy teleprompters, MASS, MetaSPO.

    benjaminhan.net/posts/20260430

    #LLMs #AI #PromptEngineering #NLP #Microsoft #ICLR

    #llms #ai #promptengineering #nlp #microsoft #iclr
  25. Benjamin Han @[email protected] ·

    SelfReflect measures whether an LLM's text summary of its uncertainty matches its actual answer distribution. Across 20 modern models: it doesn't, unless the model sees samples of its own answers first.

    The negative result does more work than the metric itself. Fits a growing line where LLM self-reports shouldn't be trusted as introspection. Practical workaround isn't cheap: N forward passes to sample, then a summarize pass.

    benjaminhan.net/posts/20260430

    #LLMs #AI #Evaluation #Apple #ICLR

    #llms #ai #evaluation #apple #iclr
  26. Benjamin Han @[email protected] ·

    SelfReflect measures whether an LLM's text summary of its uncertainty matches its actual answer distribution. Across 20 modern models: it doesn't, unless the model sees samples of its own answers first.

    The negative result does more work than the metric itself. Fits a growing line where LLM self-reports shouldn't be trusted as introspection. Practical workaround isn't cheap: N forward passes to sample, then a summarize pass.

    benjaminhan.net/posts/20260430

    #LLMs #AI #Evaluation #Apple #ICLR

    #llms #ai #evaluation #apple #iclr
  27. Northern Kender @[email protected] ·

    Сейчас будет breaking news, срыв покровов.

    В связи с поездкой на #ICLR в Рио, вспомнил песню "На далекой Амазонке" из мультфильма «Ежик плюс черепаха» 1981 года.
    И понял, что нас десятилетиями обманывали!

    В переводе Маршака пароходы в Бразилию отправляются из Ливерпуля:

    "Из ливерпульской гавани
    Всегда по четвергам,
    Суда уходят в плаванье
    К далеким берегам.
    Плывут они в Бразилию,
    Бразилию, Бразилию.
    И я хочу в Бразилию —
    К далеким берегам!"

    Но в оригинале у Киплинга вовсе не Ливерпуль никакой, а Саутгемптон!

    "Yes, weekly from Southampton,
    Great steamers, white and gold,
    Go rolling down to Rio
    (Roll down—roll down to Rio!)
    And I'd like to roll to Rio
    Some day before I'm old!"

    А ведь #Ливерпуль и #Саутгемптон даже не близко, они вообще на разных сторонах Британии! Переводческие вольности в полный рост.

    #Бразилия #Киплинг #Маршак

    #iclr #ливерпуль #саутгемптон #бразилия #киплинг #маршак
  28. Northern Kender @[email protected] ·

    Сейчас будет breaking news, срыв покровов.

    В связи с поездкой на #ICLR в Рио, вспомнил песню "На далекой Амазонке" из мультфильма «Ежик плюс черепаха» 1981 года.
    И понял, что нас десятилетиями обманывали!

    В переводе Маршака пароходы в Бразилию отправляются из Ливерпуля:

    "Из ливерпульской гавани
    Всегда по четвергам,
    Суда уходят в плаванье
    К далеким берегам.
    Плывут они в Бразилию,
    Бразилию, Бразилию.
    И я хочу в Бразилию —
    К далеким берегам!"

    Но в оригинале у Киплинга вовсе не Ливерпуль никакой, а Саутгемптон!

    "Yes, weekly from Southampton,
    Great steamers, white and gold,
    Go rolling down to Rio
    (Roll down—roll down to Rio!)
    And I'd like to roll to Rio
    Some day before I'm old!"

    А ведь #Ливерпуль и #Саутгемптон даже не близко, они вообще на разных сторонах Британии! Переводческие вольности в полный рост.

    #Бразилия #Киплинг #Маршак

    #iclr #ливерпуль #саутгемптон #бразилия #киплинг #маршак
  29. Northern Kender @[email protected] ·

    Сейчас будет breaking news, срыв покровов.

    В связи с поездкой на #ICLR в Рио, вспомнил песню "На далекой Амазонке" из мультфильма «Ежик плюс черепаха» 1981 года.
    И понял, что нас десятилетиями обманывали!

    В переводе Маршака пароходы в Бразилию отправляются из Ливерпуля:

    "Из ливерпульской гавани
    Всегда по четвергам,
    Суда уходят в плаванье
    К далеким берегам.
    Плывут они в Бразилию,
    Бразилию, Бразилию.
    И я хочу в Бразилию —
    К далеким берегам!"

    Но в оригинале у Киплинга вовсе не Ливерпуль никакой, а Саутгемптон!

    "Yes, weekly from Southampton,
    Great steamers, white and gold,
    Go rolling down to Rio
    (Roll down—roll down to Rio!)
    And I'd like to roll to Rio
    Some day before I'm old!"

    А ведь #Ливерпуль и #Саутгемптон даже не близко, они вообще на разных сторонах Британии! Переводческие вольности в полный рост.

    #Бразилия #Киплинг #Маршак

    #iclr #ливерпуль #саутгемптон #бразилия #киплинг #маршак
  30. Northern Kender @[email protected] ·

    Сейчас будет breaking news, срыв покровов.

    В связи с поездкой на #ICLR в Рио, вспомнил песню "На далекой Амазонке" из мультфильма «Ежик плюс черепаха» 1981 года.
    И понял, что нас десятилетиями обманывали!

    В переводе Маршака пароходы в Бразилию отправляются из Ливерпуля:

    "Из ливерпульской гавани
    Всегда по четвергам,
    Суда уходят в плаванье
    К далеким берегам.
    Плывут они в Бразилию,
    Бразилию, Бразилию.
    И я хочу в Бразилию —
    К далеким берегам!"

    Но в оригинале у Киплинга вовсе не Ливерпуль никакой, а Саутгемптон!

    "Yes, weekly from Southampton,
    Great steamers, white and gold,
    Go rolling down to Rio
    (Roll down—roll down to Rio!)
    And I'd like to roll to Rio
    Some day before I'm old!"

    А ведь #Ливерпуль и #Саутгемптон даже не близко, они вообще на разных сторонах Британии! Переводческие вольности в полный рост.

    #Бразилия #Киплинг #Маршак

    #маршак #киплинг #бразилия #саутгемптон #ливерпуль #iclr
  31. Northern Kender @[email protected] ·

    Сейчас будет breaking news, срыв покровов.

    В связи с поездкой на #ICLR в Рио, вспомнил песню "На далекой Амазонке" из мультфильма «Ежик плюс черепаха» 1981 года.
    И понял, что нас десятилетиями обманывали!

    В переводе Маршака пароходы в Бразилию отправляются из Ливерпуля:

    "Из ливерпульской гавани
    Всегда по четвергам,
    Суда уходят в плаванье
    К далеким берегам.
    Плывут они в Бразилию,
    Бразилию, Бразилию.
    И я хочу в Бразилию —
    К далеким берегам!"

    Но в оригинале у Киплинга вовсе не Ливерпуль никакой, а Саутгемптон!

    "Yes, weekly from Southampton,
    Great steamers, white and gold,
    Go rolling down to Rio
    (Roll down—roll down to Rio!)
    And I'd like to roll to Rio
    Some day before I'm old!"

    А ведь #Ливерпуль и #Саутгемптон даже не близко, они вообще на разных сторонах Британии! Переводческие вольности в полный рост.

    #Бразилия #Киплинг #Маршак

    #iclr #ливерпуль #саутгемптон #бразилия #киплинг #маршак
  32. BIFOLD Berlin Institute for the Foundations of Learning and Data @[email protected] ·

    #ICLR 2026 in Rio! Meet our scientists. 7 workshop papers and 7 posters co-authored by BIFOLD researchers will be presented. All details (Date of presentation, venue, etc): t1p.de/pn00w @[email protected] @[email protected] @[email protected] @[email protected] #MLSky #AI

    #iclr #mlsky #ai
  33. UKP Lab @[email protected] ·

    Big congratulations to all authors! 🚀

    #ICLR2026 #MachineLearning #AIResearch #RepresentationLearning #InformationRetrieval #DenseRetrieval #SelfSupervisedLearning #LanguageModels #NLP #UKPLab #ICLR

    #iclr2026 #machinelearning #airesearch #representationlearning #informationretrieval #denseretrieval
  34. UKP Lab @[email protected] ·

    Big congratulations to all authors! 🚀

    #ICLR2026 #MachineLearning #AIResearch #RepresentationLearning #InformationRetrieval #DenseRetrieval #SelfSupervisedLearning #LanguageModels #NLP #UKPLab #ICLR

    #iclr2026 #machinelearning #airesearch #representationlearning #informationretrieval #denseretrieval
  35. UKP Lab @[email protected] ·

    Big congratulations to all authors! 🚀

    #ICLR2026 #MachineLearning #AIResearch #RepresentationLearning #InformationRetrieval #DenseRetrieval #SelfSupervisedLearning #LanguageModels #NLP #UKPLab #ICLR

    #iclr #ukplab #nlp #languagemodels #selfsupervisedlearning #denseretrieval
  36. UKP Lab @[email protected] ·

    Big congratulations to all authors! 🚀

    #ICLR2026 #MachineLearning #AIResearch #RepresentationLearning #InformationRetrieval #DenseRetrieval #SelfSupervisedLearning #LanguageModels #NLP #UKPLab #ICLR

    #iclr2026 #machinelearning #airesearch #representationlearning #informationretrieval #denseretrieval
  37. Liz Wood @[email protected] ·

    Looking to follow (and be followed!) #mlbio #bioml #biotech #bayes #bayesian #generativemodels #startup #startups #medicine #biology #representationlearning #techbio #immunology #neurips #icml #iclr #cambridge #nyc

    #mlbio #bioml #biotech #bayes #bayesian #generativemodels
  38. tech news ᳇ eicker.news @[email protected] ·

    A major #AIconference, the International Conference on Learning Representations (#ICLR), discovered that 21% of #peerreviews were fully #AIgenerated. #Researchers raised concerns about AI-generated #reviews, citing issues like #hallucinatedcitations and #vaguefeedback. Organisers will now use automated tools to assess submissions and reviews for AI use. nature.com/articles/d41586-025 #tech #media #news

    #aiconference #iclr #peerreviews #aigenerated #researchers #reviews
  39. Liz Wood @[email protected] ·

    Got invited to speak at an #iclr workshop in May and I'm over the moon.

    #ml4bio #mlbio #ml4drugdesign #ml4celldesign

    #iclr #ml4bio #mlbio #ml4drugdesign #ml4celldesign
  40. Martin Clavey @[email protected] ·

    #ICLR : Une conférence scientifique sur l’IA polluée par l’IA générative et une fuite d’API next.ink/211622/une-conference

    #iclr
  41. Jan Tobias Muehlberg @[email protected] ·

    "What can researchers do if they suspect that their manuscripts have been peer reviewed using #AI? Dozens of academics have raised concerns on social media about manuscripts and #peerreview|s submitted to the organizers of next year’s International Conference on Learning Representations (#ICLR), an annual gathering of specialists in machine learning. Among other things, they flagged hallucinated citations and suspiciously long and vague feedback on their work."

    nature.com/articles/d41586-025

    #ai #peerreview #iclr
  42. Peter Bloem @[email protected] ·

    A thought from 2016.

    It took us nine years and a bit. In retrospect, I think we held out pretty long.

    #iclr #openreview

    #iclr #openreview
  43. Hacker News @[email protected] ·

    I invented a new generative model and got accepted to ICLR

    discrete-distribution-networks

    #HackerNews #generativemodel #ICLR #invention #machinelearning #hackernews

    #hackernews #generativemodel #iclr #invention #machinelearning
  44. Olam News @[email protected] ·

    AI Scientist-v2 Passes ICLR Peer Review Workshop

    AI Scientist-v2 makes history as the first automated research system to pass ICLR peer review with above-average human scores.

    olamnews.com/technology/ai/187

    #aiscientist #artificialintelligence #iclr #peerreview
  45. Habr @[email protected] ·

    ICLR-2025: что нового в мультимодальном ранжировании

    Всем привет! Недавно мы — Алексей Спасёнов и Кирилл Никоров из ML‑команды Мультимедиа Поиска Яндекса (Картинки и Видео) — и ещё 90 яндексоидов побывали на конференции ICLR-2025 и рады рассказать о свежих направлениях исследований в области мультимодального ранжирования. С каждым годом эта область ML приобретает всё более важной. Люди всё чаще предпочитают получать информацию из визуальных медиа и кратких ИИ‑выжимок, а не привычных текстовых статей. При этом область мультимодального ранжирования является довольно сложной и интересной, так как она постоянно использует разные данные (тексты, аудио, изображения) и требует высокой вычислительной эффективности. В этой статье мы хотели бы поделиться самыми интересными и перспективными для нашей области работами, которые мы выделили на конференции.

    habr.com/ru/companies/yandex/a

    #machinelearning #нейросети #машинное_обучение #мультимодальное_ранжирование #искусственный_интеллект #iclr #конференция #статьи

    #статьи #конференция #iclr #искусственный_интеллект #мультимодальное_ранжирование #машинное_обучение
  46. Peter Bloem @[email protected] ·

    What's going on with this #ICLR paper?

    The metareview says that the authors provided a sound rebuttal and update to the paper, but neither are available (rebuttals are shown on other papers).

    openreview.net/forum?id=2jf5x5

    #iclr
  47. Emmie Hine @[email protected] ·

    📬 Today in the WR: “#AI scientists” debut at the #ICLR conference, China launches fast EV charging and deepfake labeling, and robots go to war (not #LAWs, but close). Then, why a secretive ultramarathon made me optimistic about Bluesky’s future 🦋
    Link ⬇️

    #ai #iclr #laws
  48. Gaël Varoquaux @[email protected] ·

    🎓Paper time!✨ #ICLR spotlight. Concluding of 5 years of research on missing values handling for prediction: Beware of diminishing returns in imputation for prediction. 1/8

    #iclr
  49. José Oramas @[email protected] ·

    Go #sqIRL!.... we got two papers accepted at #ICLR 2025
    The first for our lab. Congrats to Benjamin and Thomas and all the amazing collaborators that contributed to these efforts. More details will follow soon. Too happy to type :D

    #sqirl #iclr
  50. tech news ᳇ eicker.news @[email protected] ·

    A major #AIconference, the International Conference on Learning Representations (#ICLR), discovered that 21% of #peerreviews were fully #AIgenerated. #Researchers raised concerns about AI-generated #reviews, citing issues like #hallucinatedcitations and #vaguefeedback. Organisers will now use automated tools to assess submissions and reviews for AI use. nature.com/articles/d41586-025 #tech #media #news

    #aiconference #iclr #peerreviews #aigenerated #researchers #reviews