feat: add --provider-only flag to pin OpenRouter routing

fix: compare INTEGER answers by value, not string

Update benchmark docs and parser question types

fix: use tqdm.write() for question result lines

fix: add MCQ_MATCHING and INTEGER_2 to prompt templates

fix: correct JEE Advanced marking scheme (3 bugs)

ux: sticky progress bar and live cost tracking

ux: cleaner progress bars with live score line

perf: filter records before Dataset.from_list to avoid decoding images

fix: tolerate provider-specific wrapper tokens and bare answers

feat: extend reasoning_effort to qwen/* and z-ai/*; add Qwen3-VL and GLM-4.6V

feat: sort predictions.jsonl and summary.jsonl by question_id at run end

feat: enable thinking effort for Google Gemini models (default high)

feat: enable extended thinking for Claude models (anthropic/*)

fix: write exam_name/exam_year to summary.jsonl; accumulate reprompt cost+tokens

feat: concurrent API calls via ThreadPoolExecutor (default 4 workers)

fix: show partial-credit answers as yellow ~ in progress bar, not red incorrect

fix: encode question images as PNG (lossless) instead of JPEG

fix: failure_unexpected_type scores 0 not -1

Fix remaining old model references in CLAUDE.md, README, llm_interface.py

Fix resume summary bug and add --score-only mode

Log response metadata (tokens, cost, latency) from OpenRouter API

Add reproducibility controls, leaderboard, and contamination disclosure

Simplify prompts for thinking models

Move metadata.jsonl into images/ so ImageFolder picks up metadata

Switch to ImageFolder format for HuggingFace dataset viewer

Move metadata.jsonl to repo root for HuggingFace image resolution

Add missing paper_id field to dataset features

Rename image_path to image for HuggingFace dataset viewer compatibility

Serialize correct_answer as JSON string in metadata.jsonl

Replace dataset loading script with direct JSONL loading

Fix low-priority issues: dead code, parsing, deps, tests

Fix 7 medium-priority issues: resume, timeout, logging, prompts, dead code

Fix 5 high-priority bugs: caching, rate-limit retry, scoring, types, error msg

Refactor benchmark_runner.py: extract process_question() and log_question_result() helpers

Integer type question evaluation fixed

Separation of response and score in predictions.jsonl and summary.jsonl

With results of Jee Advanced 2024

--question-ids working and force redownloading

JEE Advance 2024 added

Refactor: Generalize benchmark for JEE and NEET exams

Huggingface dataset compatibility

Improved huggingface compatability

With more clear summary report and reattempt mechanism

First working version