Jun 15

1. 18:21

   Revert anti-bleed prompt mitigation (ineffective)

   fabrizziomcl
   Verified signature

   88498da
2. 18:07

   Anti-bleed + confusion-recovery guidance when no reading passage

   fabrizziomcl
   Verified signature

   58562ec
3. 18:01

   Add LLM-judge (MRBench) socratic-quality results

   fabrizziomcl
   Verified signature

   c18c9a9
4. 17:42

   Update README: bilingual model + rigorous eval (capability + behavior)

   fabrizziomcl
   Verified signature

   ce8308a
5. 17:21

   Switch to bilingual LoRA model (educrate-qwen3-bi); hide <think>; EN+ES support

   fabrizziomcl
   Verified signature

   496a1bd
6. 10:15

   Point Space to org-hosted model build-small-hackathon/educrate-qwen3-sft

   fabrizziomcl
   Verified signature

   83ebd41
7. 10:13

   remove stray pyc

   fabrizziomcl
   Verified signature

   de24331
8. 10:05

   Professional minimalist UI (English, no emojis); SFT model; tighter decoding + anti-hallucination prompt

   fabrizziomcl
   Verified signature

   89447c9
9. 09:58

   Add video/post link placeholders

   fabrizziomcl
   Verified signature

   be2d065
10. 09:56

    Add hackathon tags (track:backyard, badges) + write-up

    fabrizziomcl
    Verified signature

    63ec228
11. 09:53

    EduCrate — Spanish Socratic tutor (Qwen3-0.6B SFT+GRPO), runs on CPU

    fabrizziomcl
    Verified signature

    82d5c6f
12. 09:53

    initial commit

    fabrizziomcl
    Verified signature

    3b3c461