Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

GPT-5.2

AI overview

Sign in and the AI will write an overview from our coverage.

Headlines · 5

  1. Essay argues AI personality is a post-training strategy

    A Juejin translation of Sean Goedecke's essay argues that giving language models a coherent personality during post-training can guide useful behavior, consistency and safer outputs. It describes personality as an engineering term rather than evidence of sentience, while acknowledging the counterargument that human-like framing may encourage users to overestimate model capabilities. The article offers no independent measurements for its stronger claims about capability gains or losses.

    掘金 · 人工智能 · 🔥 0
  2. SciMuse puts AI-generated research ideas before expert reviewers

    A PNAS study introduced SciMuse, a system that combines a knowledge graph spanning 58 million papers with language models to generate personalized interdisciplinary research ideas. Max Planck research group leaders submitted 4,451 valid blind ratings, with proposals averaging 2.40 out of 5 for interest and 24.9% receiving scores of 4 or 5. In a separate comparison, 32.5% of GPT-5.2 proposals received high scores, compared with 29.1% for proposals from senior physics doctoral students and 25.0% for GPT-4 proposals.

    虎嗅 AI · 🔥 0
  3. OpenAI and Ligo launch BioBasicsBench for wet-lab analysis

    OpenAI introduced BioBasicsBench, a 100-task benchmark developed with Ligo to evaluate whether AI agents can interpret raw experimental data, identify errors and recognize when evidence cannot support a conclusion. Tasks cover protein methods, imaging and cytometry, quantification and assays, nucleic acid methods and molecular cloning. Scientists built the tasks from their own experimental data and paired them with expert-written grading criteria. OpenAI reported that GPT-6 Astra earned a 66.7% mean reward at xhigh reasoning effort, compared with 43.8% for GPT-5.6 Sol.

    OpenAI · 🔥 0
  4. LifeVal tests personalized AI assistants on everyday tasks

    OpenAI introduced LifeVal, a benchmark for evaluating personalized AI assistants on everyday consumer tasks involving browsing and computer use. The reports conflict on whether it contains 148 or 169 tasks. One report says that, among conditions requiring information from prior ChatGPT conversations, the average success rate was 33.33%. It also says that moving background information into conversation history reduced the relevant rate from 55.8% to 41.5%. The paper reportedly does not identify which models were used for the personalization analysis. LifeVal’s paper and results were released, but its full task dataset and evaluation code were not.

    iThome 台湾 · 🔥 0
  5. Ai2 benchmark exposes brittleness in LLM research-idea novelty judging

    Ai2 released the study “Old Ideas, Novel Problems,” along with a benchmark, data, code, and an automated pipeline for building novelty-judge evaluation sets. Across six LLM judges, small changes to prompts and evaluation setups produced sharply different verdicts; in one test, GPT-5.4 accuracy fell from 93.5% to 40.9% after a prompt change, while retrieval and additional reasoning did not consistently improve results. On one Human+Generated test set, a simple Claude Opus 4.6 setup achieved 0.854 macro-F1 at a reported cost of $2.59, outperforming two dedicated evaluators in the study’s reported comparison.

    Ai2 Blog · 🔥 0

Experience and discussion from the community

Share my GPT-5.2 experienceAsk about GPT-5.2

MIRIX-2.0: Evidence Fidelity, Experience Evolution

The most direct way to give an Agent long-term memory is to write history into a database and retrieve a few relevant records when a new question arrives. This pipeline soon encounters two different requirements: a personal assistant needs to remember specific people, events, times, and preferences; a task Agent needs to learn from past actions what to do next. The former fears losing details to c…

MIRIXMemory元学习HypergraphAutoDream

arxiv.orghttps://arxiv.org/abs/2611.12068