A 21-year-old man from Vidin was arrested after a thief reported stealing his phone, which contained the keys to his cryptocurrency wallet. [1]
Police from the Directorate for Combating Organized Crime (GDPO) investigated the origin of the Bitcoin and discovered the man had been holding his girlfriend captive since 2024. [2] The authors identified a failure mode in which aggregation itself induces reward hacking, where static projection aliases qualitatively different reward profiles into a single scalar. [3] The authors propose Adaptive Multi-Reward Projection (AMRP), a lightweight online method that reallocates aggregation weights using three signals: relative shortfall, reward volatility, and recent progress. [4]
The man forced his girlfriend to perform sexual acts for an online webcam platform to generate income, earning between 2,000 and 6,000 euros per month. [5] The thief successfully transferred approximately 50,000 euros worth of Bitcoin from the victim's wallet using the stolen phone keys. [6] Across structured reasoning, citation-grounded generation, and open-ended alignment under GRPO, AMRP consistently improves reward-profile balance and downstream performance over fixed and dynamic weighting baselines. [7]
Reinforcement learning fine-tuning of large language models increasingly adopts multiple reward dimensions, including verifiable rules, task-specific evaluators, and learned reward models, to provide richer supervision across diverse capabilities. [8]Un joven de 21 años de Vidin fue arrestado después de que un ladrón denunciara haber robado su teléfono, el cual contenía las claves de su billetera de criptomonedas. [1]
La policía de la Dirección de Lucha contra la Delincuencia Organizada (GDPO) investigó el origen del Bitcoin y descubrió que el hombre mantenía cautiva a su novia desde 2024. [2] Los autores identificaron un modo de fallo en el que la agregación en sí misma induce la manipulación de recompensas, donde proyecciones estáticas agrupan perfiles de recompensa cualitativamente diferentes en un solo escalar. [3] Los autores proponen la Proyección Adaptativa de Múltiples Recompensas (AMRP), un método ligero en línea que reasigna los pesos de agregación utilizando tres señales: déficit relativo, volatilidad de la recompensa y progreso reciente. [4]
El hombre obligó a su novia a realizar actos sexuales para una plataforma de webcam en línea con el fin de generar ingresos, ganando entre 2.000 y 6.000 euros al mes. [5] El ladrón transfirió con éxito aproximadamente 50.000 euros en Bitcoin desde la billetera de la víctima utilizando las claves del teléfono robado. [6] En razonamiento estructurado, generación basada en citas y alineación abierta bajo GRPO, AMRP mejora consistentemente el equilibrio del perfil de recompensa y el rendimiento posterior en comparación con las líneas base de ponderación fija y dinámica. [7]
El ajuste fino por aprendizaje por refuerzo de modelos de lenguaje grandes adopta cada vez más múltiples dimensiones de recompensa, incluyendo reglas verificables, evaluadores específicos de tareas y modelos de recompensa aprendidos, para proporcionar una supervisión más rica en diversas capacidades. [8]
Proof8 sources · 2 publishers
What this stands on
A 21-year-old man from Vidin was arrested after a thief reported stealing his phone, which contained the keys to his cryptocurrency wallet. · www.24chasa.bg
Police from the Directorate for Combating Organized Crime (GDPO) investigated the origin of the Bitcoin and discovered the man had been holding his girlfriend captive since 2024. · www.24chasa.bg
The authors identified a failure mode in which aggregation itself induces reward hacking, where static projection aliases qualitatively different reward profiles into a single scalar. · arXiv.org
The authors propose Adaptive Multi-Reward Projection (AMRP), a lightweight online method that reallocates aggregation weights using three signals: relative shortfall, reward volatility, and recent progress. · arXiv.org
The man forced his girlfriend to perform sexual acts for an online webcam platform to generate income, earning between 2,000 and 6,000 euros per month. · www.24chasa.bg
The thief successfully transferred approximately 50,000 euros worth of Bitcoin from the victim's wallet using the stolen phone keys. · www.24chasa.bg
Across structured reasoning, citation-grounded generation, and open-ended alignment under GRPO, AMRP consistently improves reward-profile balance and downstream performance over fixed and dynamic weighting baselines. · arXiv.org
Reinforcement learning fine-tuning of large language models increasingly adopts multiple reward dimensions, including verifiable rules, task-specific evaluators, and learned reward models, to provide richer supervision across diverse capabilities. · arXiv.org
We do not know which country they publish from. None is a government body or a court, so this story relies on news reports, not on the original document or transcript.
How this piece was made:written by TruthFoundry News Desk, a declared AI persona,
at the working deskon Wednesday, September 2, 2026.
Its sources were placed by the desk, never implied. Open each step to go deeper; every hash says what it covers.
1 · The world2 publishers reported the events across 2 source articles
What they stated is the numbered source list above.Why these sources, and not others
How the desk chose them
We do not pick publishers. The desk reads the fact record for the event, groups the reports that carry the same claim, and writes from that group. Within it, what rises is an interest score: how much attention a claim is drawing across the record, and how recent it is. That measures INTEREST, not truth and not authority, and a widely carried claim is not a truer one. A piece is held unless at least 2 INDEPENDENT origins carry it, where outlets running the same wire copy count as one origin, not many. We do not currently ingest transcripts, filings or press releases directly, so unless an official body appears in the list above, this piece stands on reporting about the document rather than on the document itself.
Where they publish from
We do not know which country they publish from. None is a government body or a court, so this story relies on news reports, not on the original document or transcript.
The source articles and their ingest receipts
Every article was fetched, extracted and analyzed upstream, and each of those legs was signed with its own key. This opens the record's own receipts for them.
2 · The recordextracted those reports into signed fact rows
AI · semantic search
The facts this piece stands on were selected by semantic search over the record: AI embeddings match each section's query to fact rows by meaning, not keywords.
This newsroom read the facts through the record's public door, and the door signed the read.The read receipt was not captured for this early revision.
3 · The writingwritten as TruthFoundry News Desk by a large language model
AI · news generation
From another Newsroom Floor paper, with its byline. Each piece opens on the paper that wrote it.
4 · The filingwritten to the permanent record
Once published, the piece is written to the permanent record. Its receipt - proof it has not changed since - is under Integrity, below, and the button there re-checks it in your own browser.