Pinned Loading

1. devbench devbench Public

   Developer-facing benchmark for structured geospatial LLM output. Implements VOYGR's Phase 2 evaluation (Section 7, Q1 2026 report) measuring how Claude, OpenAI, and Gemini handle geocoding, GeoJSON…

   Python

2. GuardLayer GuardLayer Public

   FastAPI middleware that intercepts LLM place recommendations and verifies them against VOYGR's business-status API before they reach users — catching permanently closed venues, fabrications, and wr…

   Python

3. InCoRes InCoRes Public

   Implementation of CorefInst (TACL 2026) for Hindi, Tamil, and Bengali using the TransMuCoRes dataset. The model fine-tunes a decoder-only LLM (Llama 3.1 8B) with QLoRA to predict coreference cluste…

   Python

4. context-decay context-decay Public

   Measuring instruction-following collapse in small local LLMs — how far a constraint can sit from the query before it's forgotten, and how quantization, prompt format, and conflicting instructions m…

   Python

5. log-analytics-engine log-analytics-engine Public

   A high-throughput Apache CLF log processing pipeline that parses and aggregates server logs significantly faster than pandas, using Numba for parallel byte-level parsing and JAX for vectorised aggr…

   Python

6. claude-forge claude-forge Public

   A token-efficient orchestration framework for Claude Code. Uses Claude Opus for high-level planning/review while executing parallel steps via cheaper models, enforced by strict context and schema c…

   Python