Commit History

  1. Add FinCalc-NLI eval (financial arithmetic) + results for all checkpoints⁠

    Verified signature
    b639183

  1. add qwen3.5-4b-nli-v5-nvfp4 (ModelOpt NVFP4 PTQ + QAD) with SGLang bench

    Verified signature
    058a6c2

  1. card: links to the 4B / 2B / 0.8B v5 checkpoints

    Verified signature
    552759d
  2. add qwen3.5-0.8b-nli-v5

    Verified signature
    c64009c
  3. add qwen3.5-2b-nli-v5

    Verified signature
    dfdc1cb
  4. card: openjev-4B v5 (JevBench 0.814 public, judging/faithfulness/IF tables, determinism, injection limits, contamination manifest)

    Verified signature
    e9b1201
  5. add contamination manifest (train+test splits used for v5 panel)

    Verified signature
    7ac2931
  6. add RESULTS-v5.md

    Verified signature
    c59abe0
  7. code: gen_ifeval.py

    Verified signature
    e368d07
  8. code: think_probe.py

    Verified signature
    3eced74
  9. code: eval.py

    Verified signature
    b35f155
  10. code: train.py

    Verified signature
    b6aa6e6
  11. code: distill_teacher.py

    Verified signature
    427c2b8
  12. code: hard_gen.py

    Verified signature
    a9bf529
  13. code: data_mix.py

    Verified signature
    b5bd9a2
  14. code: webql_null.py

    Verified signature
    1ccf050
  15. code: eval_agentic.py

    Verified signature
    f2e1333
  16. code: eval_extra.py

    Verified signature
    3909a82
  17. code: eval_security.py

    Verified signature
    b390a79
  18. code: order_test.py

    Verified signature
    8e01204
  19. code: eval_jevbench.py

    Verified signature
    be3484d
  20. code: modeling_openjev.py

    Verified signature
    106dab5
  21. code: openjev_decide.py

    Verified signature
    4d701a9
  22. 4B v5: typed-decision cross-encoder (panel + gpt-5.5 hard items + teacher distillation)

    Verified signature
    5df9876

  1. SGLang serving: external model package, launcher, client, benchmark; card section

    Verified signature
    f004f37

  1. Document the 0.8B v2s-long checkpoint served by the Space

    Verified signature
    4395b29
  2. upload openjev-0.8B v2s-long (stage3, 4k ctx, faith/ifollow/false-premise + long-doc)

    Verified signature
    0271e1a
  3. Share Qwen3.5 prefix cache across NLI hypotheses (#1)⁠

    AlexWortega +1: epsilon3
    65ae5e7

  1. upload qwen3.5-35b-a3b-nli

    Verified signature
    4b5f9a6
  2. upload qwen3.5-4b-nli-v2

    Verified signature
    9f176ba
  3. openjev v2: results, videos (Doom v2, Minecraft), code, card

    Verified signature
    0988d66

  1. add results/webql_geminiprompt_4b.json

    Verified signature
    f8187e6
  2. add results/webql_bench_4b.json

    Verified signature
    466a519
  3. add results/webql_scaffold_4b.json

    Verified signature
    7c326a9
  4. add results/webql_mlp_4b.json

    Verified signature
    9d9aabe
  5. add results/webql_fulldoc_35b.json

    Verified signature
    139fb28
  6. add results/webql_mlp_35b.json

    Verified signature
    c113a3e
  7. add results/webql_bench_35b.json

    Verified signature
    90d97ca
  8. add results/webql_fulldoc_4b.json

    Verified signature
    06d1cbe
  9. add results/webql_geminiprompt_35b.json

    Verified signature
    4473830
  10. add results/latent_mlp_35b.json

    Verified signature
    3d5cbef
  11. add results/qwen35b_all.json

    Verified signature
    9cba221
  12. add assets/radar_openjev_35b.png

    Verified signature
    d785095
  13. add code/webql_gemini_prompt.py

    Verified signature
    5a0cd95
  14. add code/webql_fulldoc.py

    Verified signature
    ed99575
  15. add code/webql_mlp.py

    Verified signature
    7f48358
  16. add code/webql_scaffold.py

    Verified signature
    6053aa9
  17. add code/webql_bench.py

    Verified signature
    8ce2faf
  18. add code/latent_mlp.py

    Verified signature
    0027527
  19. add code/radar35.py

    Verified signature
    18c48ba